Alternatives to Gemma 4

GPT-6.1 Sol is an upgraded OpenAI model that combines advanced intelligence with lower operating costs for coding, professional knowledge work, computer use, scientific research, and autonomous agents. OpenAI positions it as offering near-GPT-6 Astra intelligence at one-fifth of Astra's standard input and output token prices. The model delivers substantial improvements over GPT-6 Sol in software engineering, complex document understanding, business automation, and long-horizon computer-use workflows. On DeepSWE v1.1, GPT-6.1 Sol matches GPT-6 Astra at approximately one-fifth of the cost and surpasses GPT-6 Sol's highest score by 6.4 percentage points at lower reasoning effort. On AutomationBench, it scores 4.8 percentage points higher than GPT-6 Sol at the same reasoning setting and 2.2 points above Opus 5.5 at medium reasoning effort. GPT-6.1 Sol also improves computer use, coming within 2.1 percentage points of GPT-6 Astra on the OSWorld 2.0 offline set at maximum reasoning effort while costing roughly one-seventh as much per task. For scientific workflows, the model more than doubles GPT-6 Sol's Terminal-Bench Science 0.1 score at maximum effort while reducing average task cost by more than half. Factuality has also improved, with the share of responses containing a factual error at low reasoning effort falling from 11.4% with GPT-6 Sol to 7.7% with GPT-6.1 Sol on OpenAI's difficult error-focused evaluation. Developers can access GPT-6.1 Sol through the OpenAI API for $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens, while eligible users can access it through ChatGPT Work and Codex.

GPT-6.1 Terra is a multimodal reasoning and computer-use model developed by OpenAI to support visual analysis, software engineering, writing, and automated professional workflows. The model processes text and image inputs and generates text outputs, enabling applications that require understanding visual information and executing actions in software interfaces. Its computer-use capabilities allow AI agents to interact with desktop applications, complete forms, interpret technical drawings, and perform multistep tasks across graphical environments. GPT-6.1 Terra provides a 1,050,000-token context window and supports outputs of up to 128,000 tokens for processing extensive documents and complex task instructions. Its reasoning effort can be configured across none, low, medium, high, xhigh, and max settings to balance computational effort, response speed, and task complexity. Software development capabilities include code generation, debugging, command-line problem-solving, and automated execution of technical workflows. The model supports function calling, structured outputs, web search, file search, code interpretation, computer use, and Model Context Protocol integrations through the OpenAI Responses API. Compared with GPT-5.6 Terra, the model provides improved visual reasoning, coding performance, factual accuracy, and computer-use capabilities while reducing standard API token prices by 50%. GPT-6.1 Terra is available in ChatGPT Work, Codex, and the OpenAI API, with standard pricing of $1 per million input tokens and $5 per million output tokens.

GPT-6 Astra is an advanced general-purpose AI model from OpenAI designed for demanding agentic, technical, scientific, and professional workflows. The model combines reasoning with computer-use capabilities that let it interact with applications, websites, development environments, productivity software, and specialized tools. It can perform activities such as online research, CRM updates, form completion, calendar organization, data analysis, software testing, troubleshooting, and frontend quality assurance. Astra is also trained for professional artifact creation, including documents, presentations, spreadsheets, analyses, websites, applications, and visual designs that follow provided templates and business conventions. Its software engineering capabilities support complex coding, codebase understanding, debugging, database migration, verification, and extended autonomous development tasks. OpenAI has also introduced a Codex context system for Astra that can maintain notes across context windows and search earlier conversation and tool history instead of relying exclusively on repeated summarization. Scientific capabilities span areas such as mathematics, biology, chemistry, health, data analysis, and the use of specialized research software. The model includes strengthened alignment and cybersecurity safeguards intended to help it remain within authorized task boundaries while supporting legitimate activities such as secure code review and vulnerability remediation. GPT-6 Astra is available across supported ChatGPT plans and developer platforms, including the OpenAI API and Amazon Web Services infrastructure.

GPT-5.6 Sol is OpenAI’s flagship model in the GPT-5.6 series, built for high-end reasoning, coding, scientific analysis, cybersecurity, and agentic automation. The model is designed to handle complex tasks that require planning, iteration, tool coordination, long-horizon reasoning, and careful execution across multiple steps. GPT-5.6 Sol introduces max reasoning effort, giving the model more time to reason deeply through difficult problems. It also introduces ultra mode, which uses subagents to accelerate complex work and extend capability beyond a single-agent workflow. For coding, GPT-5.6 Sol is positioned for command-line workflows, software engineering tasks, debugging, testing, and multi-step tool use. In biology and quantitative research workflows, the model is designed to support genomics analysis and other long-context scientific tasks while using tokens more efficiently than prior models. For cybersecurity, GPT-5.6 Sol supports legitimate defensive work such as vulnerability research, code review, patch development, security education, and defensive testing. The model includes a layered safeguard stack with trained refusals, real-time cyber and biology misuse classifiers, account-level monitoring, differentiated access, human-in-the-loop review, and ongoing red-team testing. GPT-5.6 Sol helps trusted users and organizations access more powerful AI for technical work while maintaining stronger controls around misuse, sensitive requests, and high-risk activity.

GPT-6 Sol is a frontier AI model from OpenAI built for professional knowledge work, software development, business automation, computer use, and agentic applications. It brings many of the advances introduced with GPT-6 Astra to a model optimized for greater cost efficiency and higher-volume use. GPT-6 Sol can apply configurable reasoning effort to complex tasks, giving developers and users control over how much computation is used for different workloads. Its coding capabilities support long-horizon software engineering, codebase modification, debugging, testing, and agent-driven development in tools such as Codex. For professional workflows, the model can work across applications and tools to complete multi-step tasks in areas such as sales, marketing, operations, finance, support, and HR. OpenAI also reports improved factuality compared with GPT-5.6 Sol, reducing errors on its internal evaluation of difficult real-world conversations. Computer-use capabilities allow Sol-powered agents to navigate software interfaces and perform extended workflows that require multiple actions and decisions. Enhanced prompt caching provides higher cache-hit rates and discounts on reused input context, making the model better suited to persistent agents and long conversations. GPT-6 Sol is available to developers through the OpenAI API as gpt-6-sol and is also offered in ChatGPT Work and Codex.

GPT-6 Luna is a lightweight, cost-efficient model in OpenAI’s GPT-6 family built for coding, professional tasks, computer use, and high-volume AI applications. The model incorporates advances from the same generation as GPT-6 Astra while emphasizing lower inference costs and greater efficiency for everyday workloads. API pricing is $0.10 per million input tokens and $0.50 per million output tokens, making Luna suitable for applications that process large volumes of requests. In professional work, GPT-6 Luna can execute multi-step workflows involving business applications, tools, and structured tasks across functions such as sales, marketing, operations, support, finance, and HR. For software engineering, the model can work on real codebases, perform extended development tasks, and operate within coding agents such as Codex. Its computer-use capabilities allow AI agents to interact with software interfaces and carry out longer workflows that require repeated actions and decisions. OpenAI also reports substantial factuality improvements over GPT-5.6 Luna, with higher reasoning settings enabling stronger performance on difficult factual questions. GPT-6 prompt caching provides higher cache-hit rates and discounted cached input, helping persistent agents and long conversations reuse context more efficiently. GPT-6 Luna is available through ChatGPT Work, Codex, the OpenAI API, and the ChatGPT desktop app for eligible users.

Grok 4.6 is an advanced xAI model built for long-running agents, coding, knowledge work, interactive applications, and visual project creation. It improves on Grok 4.5 with a focus on staying with complex tasks across many steps, whether the user is researching a topic, analyzing information, working across a codebase, or building a polished application. The model was trained through a longer supplemental run that used curated model-generated reasoning data, advanced technical concepts, engineering data, and an improved training recipe. Grok 4.6 was also trained with SFT and RL across domains such as STEM, software engineering, knowledge work, kernel optimization, web development, computer-aided design, and agentic coding. It is designed to turn ambitious ideas into working projects by researching unfamiliar domains, defining application structure, building core interactions, and iterating through feedback. The model shows stronger first passes on visual and interactive projects, helping users establish structure and visual language more quickly. Grok 4.6 also demonstrates more self-testing and verification during longer trajectories. It is available through Cursor, Grok Build, the xAI API, OpenRouter, Vercel, Cloudflare, and other partners, with pricing starting at $2 per million input tokens and $6 per million output tokens. By combining frontier reasoning, agentic coding, long-running task execution, visual project generation, API access, and broad developer availability, Grok 4.6 helps builders move from idea to working software faster.

Grok 4.7 is SpaceXAI’s advanced AI model for software development, professional knowledge work, and multi-step agentic tasks. It is built on a larger base model than Grok 4.6 and uses a longer reinforcement learning training run weighted toward difficult problems that require extended execution time. The model is better at checking its own work, maintaining longer context, and completing complex workflows that span many steps. Native support for the Grok Bot harness improves its conversational abilities and makes it more capable across general knowledge and interactive work. Grok 4.7 is positioned for use in coding, terminal automation, legal workflows, document and presentation creation, electrical engineering, healthcare reasoning, and other professional tasks. SpaceXAI also reports improvements across software engineering benchmarks including CursorBench and DeepSWE as well as professional-work evaluations such as AA Briefcase. The model includes a redesigned safety system intended to improve refusal behavior, jailbreak resistance, and handling of potentially dangerous cybersecurity and biological requests. Developers can access Grok 4.7 through Grok Build, Cursor, the Grok API, third-party coding tools, cloud providers, and model-routing platforms. Grok 4.7 is priced starting at $2 per million input tokens and $6 per million output tokens, with a faster variant available at a higher price.

Muse Spark Open is an open-weight foundation model developed by Meta to support multimodal reasoning, autonomous agents, and complex software development tasks. The model features 816 billion total parameters with approximately 42 billion active parameters and a context window supporting up to one million tokens. Its architecture is designed to process multimodal information while maintaining context across lengthy workflows, extensive codebases, and multistep assignments. Muse Spark Open provides capabilities for software engineering, code generation, debugging, application development, and automated testing. Its agentic functionality supports external tool interactions, task planning, computer-based workflows, and execution of complex objectives over extended sessions. The model can run locally using two Apple M5 Ultra Mac Studio systems with the MLX framework, enabling organizations to process information without relying on hosted inference services. Developers can deploy the model through SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, llama.cpp, and other supported inference engines. Customization is supported through PyTorch and TorchTitan, allowing organizations to adapt the model to specialized applications and workflows. Muse Spark Open is distributed under the Muse Model License and is available through Hugging Face, major cloud model repositories, and hosted API providers including Together AI, Fireworks AI, and OpenRouter.

Kimi K3.1 Pro is an advanced multimodal AI reasoning model developed by Moonshot AI for software engineering, autonomous agents, technical design, and analytical knowledge work. Built on the Kimi K3 architecture, the model uses approximately 2.8 trillion total parameters with 104 billion active parameters and supports a one-million-token context window. Its enhanced reasoning capabilities enable complex, multistep problem-solving across programming, scientific analysis, financial research, and business operations. Coding features include source code generation, debugging, repository-level modifications, automated testing, and end-to-end application development. The model can create interactive websites, Three.js environments, games, and software projects using development frameworks and engines such as Unity, Unreal Engine, and Godot. Its spatial reasoning capabilities support CAD code generation, parametric modeling, 3D asset creation, and geometry adjustments through tools such as Blender. Financial analysis capabilities include spreadsheet modeling, document-based research, numerical reasoning, and investment analysis using information from multiple sources. Kimi K3.1 Pro supports text and image inputs, external tool use, and long-running agentic workflows that require planning and iterative execution. The model is accessible through Kimi applications and APIs, with an additional Highspeed version designed to deliver approximately two to three times faster responses.

NVIDIA Nemotron 3.5 Super VL is a multimodal foundation model developed for AI agents that require reasoning, visual understanding, coding, and multistep task execution. Its 120-billion-parameter mixture-of-experts architecture activates approximately 12 billion parameters per token, enabling efficient inference while maintaining substantial model capacity. The model processes text, images, documents, and video within a unified system that combines multimodal interpretation with instruction following and tool use. Its document intelligence capabilities include extracting information from tables, analyzing charts, interpreting page layouts, and processing multipage documents. Video understanding features support visual information retrieval, content review, incident analysis, and reasoning across recorded sequences. The model also supports software engineering tasks such as code analysis, error identification, debugging, testing, and recommending fixes. Developers can customize Nemotron 3.5 Super VL using NVIDIA NeMo for supervised fine-tuning, low-rank adaptation, and reinforcement learning with domain-specific data. Its inference optimizations include speculative decoding, context reuse, multi-token prediction, and NVFP4 quantization to improve throughput and reduce computational requirements. The model is available with open weights and supports deployment across compatible local hardware, enterprise data centers, and cloud infrastructure.

NVIDIA Nemotron 3.5 Bolt is a series of efficient AI models developed by NVIDIA for running autonomous agents and conversational applications directly on local hardware. The family consists of Bolt 74B-A9B 24GB and Bolt E4B 4GB, which address different performance and memory requirements. Bolt 74B-A9B uses a mixture-of-experts architecture with 74 billion total parameters and approximately 9 billion active parameters to balance reasoning capabilities with computational efficiency. Native 2-bit quantization enables the model to operate within a 24GB memory footprint on supported NVIDIA systems. Its agentic capabilities include web research, code generation, software debugging, information retrieval, and multistep task execution. The model can be integrated with local AI agent frameworks and development tools to complete assignments without sending application workloads to external inference services. Bolt E4B is optimized for fast conversational responses, multistep interactions, and embedded AI experiences such as virtual assistants and interactive game characters. NVIDIA supports local deployment of the Bolt family across compatible GeForce RTX, RTX PRO, DGX Spark, and Jetson hardware, depending on the model variant. Bolt 74B-A9B is available with open weights through Hugging Face and supports inference tools including llama.cpp, Ollama, LM Studio, Unsloth, and vLLM.

Grok 4.5 is SpaceXAI’s smartest model, designed to excel at coding, agentic workflows, engineering tasks, and knowledge work. The model was trained on large-scale datasets covering coding, science, engineering, and math, with additional reinforcement learning focused on multi-step software engineering. It is built to perform well on real engineering workflows, including debugging, terminal-based tasks, complex code generation, Rust and C/C++ development, and app building from minimal prompts. Grok 4.5 is served at fast-model speeds while using fewer output tokens on comparable coding tasks, helping teams complete technical work more quickly and cost-effectively. The model is also available in Grok Build, where it can help create Excel models, PowerPoint presentations, Word documents, diagrams, business review decks, and research-supported productivity assets. Developers can access Grok 4.5 through the SpaceXAI API, Cursor, and Grok Build, with simple API key setup and support for direct integration into coding and automation workflows. Its pricing is positioned for high-intelligence work at scale, with per-million-token rates for both input and output usage. Grok 4.5 is also trained for agentic execution, allowing it to handle longer technical rollouts and multi-step problem solving more effectively. For developers, engineering teams, and knowledge workers, Grok 4.5 provides a powerful AI model for software creation, office automation, technical reasoning, and production-grade agent workflows.

Muse Wonder is an advanced AI foundation model developed by Meta to support personal AI agents, professional knowledge work, and software engineering tasks. The model is part of the Muse Wonder series, which includes Muse Wonder and the larger Muse Wonder Full variant. Its Active Memory capability maintains important contextual information across conversations and retrieves relevant details when they become useful for subsequent tasks. Active Focus enables the model to pursue multistep objectives, evaluate alternative approaches, respond to obstacles, and continue working toward completion. Muse Wonder supports applications involving legal document analysis, financial information, health-related questions, and business workflow automation. Its coding capabilities include generating applications, modifying existing codebases, tracking project requirements, and iterating through testing and debugging cycles. The model can operate as part of agentic systems that interact with browsers, external applications, and connected tools to complete user-directed activities. Meta has evaluated Muse Wonder on benchmarks covering personal assistance, professional agent tasks, long-term memory, sustained execution, and software development. Muse Wonder is accessible through the Meta Model API and Muse Code, with plans to incorporate its capabilities into the Muse assistant and additional AI-powered experiences.

GLM-5.2 is a next-generation large language model built for users who need strong reasoning, coding support, and agentic AI capabilities. It can assist with complex software development tasks, technical problem-solving, automation workflows, and advanced research projects. The model is designed to process long-context information, which makes it helpful for analyzing large documents, reviewing codebases, and maintaining continuity across multi-step tasks. GLM-5.2 supports developers and organizations that want to create AI-powered tools capable of planning, reasoning, and executing more sophisticated workflows. Its architecture is structured to deliver high performance while improving efficiency for demanding AI use cases. Businesses can use GLM-5.2 to enhance productivity, streamline engineering processes, and build more capable intelligent applications. It is also useful for teams that need AI assistance across documentation, data interpretation, coding, testing, and workflow automation. The model’s emphasis on agentic engineering makes it well-suited for applications that require more than simple text generation. GLM-5.2 provides a flexible AI foundation for companies looking to bring advanced reasoning and automation into their products or internal operations.

Kimi K3 is a large-scale AI model from Moonshot AI designed for advanced reasoning, software engineering, visual understanding, agentic workflows, and knowledge work. The model is built with 2.8 trillion parameters and uses Kimi Delta Attention, a hybrid linear attention design created to support long-context intelligence. It also includes Attention Residuals and a native 1 million token context window, giving developers room to work with large files, repositories, documentation sets, transcripts, and enterprise knowledge bases. Kimi K3 always runs with thinking mode enabled and currently supports maximum reasoning effort by default. Developers can access the model through Moonshot’s OpenAI-compatible API using Python, cURL, and the OpenAI SDK. The API supports standard chat completions, streaming output, structured JSON Schema responses, partial continuation from a prefix, custom tool calling, required tool choice, and dynamic tool loading. Kimi K3 also supports vision inputs, including local images encoded as base64 and video files uploaded through the file API. Automatic context caching helps repeated long-prefix workflows become more efficient without requiring manual cache IDs or extra cache parameters. By combining long context, visual understanding, tool use, structured output, and advanced reasoning, Kimi K3 is built for developers creating sophisticated AI agents, coding systems, research tools, and enterprise applications.

GPT-5.6 Luna is OpenAI’s fast, cost-efficient model in the GPT-5.6 lineup. The GPT-5.6 family includes Sol for flagship performance, Terra for balanced everyday work, and Luna for strong capability at the lowest listed price. Luna is designed for users who need scalable AI support for routine tasks, coding assistance, workflow automation, analysis, and production API use cases where speed and cost matter. According to the pasted preview text, Luna is priced below both Sol and Terra, making it the most affordable GPT-5.6 option for high-volume workloads. The model is included in GPT-5.6 benchmark previews across Terminal-Bench 2.1, GeneBench v1, ExploitBench, and ExploitGym, showing that it is part of the same technical family used for coding, biology, and cybersecurity evaluations. Luna benefits from safeguards developed across the GPT-5.6 series, including model-level refusal training, real-time cyber and biology misuse classifiers, account-level signals, differentiated access, monitoring, enforcement, and ongoing testing. These controls are designed to preserve legitimate use cases such as debugging, code review, defensive testing, security education, and productivity automation while constraining prohibited misuse. GPT-5.6 Luna is planned for broader access through ChatGPT, Codex, and the API after the limited preview period. GPT-5.6 Luna helps developers and organizations run useful AI workflows with a practical balance of affordability, responsiveness, and safety.

GLM-5.3 is Z.ai’s advanced coding and agentic reasoning model built through scaled post-training on top of the GLM-5.2 base model. The release focuses on frontier coding, long-horizon software engineering, agent tasks, cyber evaluation, and reinforcement learning at scale. GLM-5.3 improves significantly over GLM-5.2 on complex coding benchmarks, real-world engineering environments, Terminal Bench 3.0, DeepSWE, Agents’ Last Exam, and Z.ai’s internal Code Bench. The model is trained on environments that resemble real professional work, including tasks involving codebases, infrastructure, documentation, compute clusters, experiments, bottleneck diagnosis, implementation, testing, and measurable optimization. Z.ai’s post-training stack includes IndexShare for efficient long-context processing, SAO for reinforcement learning on long-horizon tasks, and slime for large-scale asynchronous training. GLM-5.3 supports three thinking effort levels, including low, high, and max, with max recommended for coding tasks. The model also demonstrates emergent cyber capabilities across vulnerability discovery and exploitation benchmarks, prompting continued safety evaluation and hardening before weights are released. GLM-5.3 can be used through the GLM Coding Plan, ZCode, Claude Code, OpenCode, and other coding agent workflows. By combining stronger coding performance, long-horizon task execution, post-training scale, cyber evaluation, reasoning effort controls, and coding-agent integrations, GLM-5.3 supports advanced developer and research workflows.

Qwen3.8-Max is a frontier AI model from the Qwen family designed for advanced coding, coworking, research, multimodal reasoning, and long-horizon autonomous tasks. It is described as Qwen’s most capable model to date and the first Qwen-Max-class model with open weights announced for release. The model scales to 2.4 trillion parameters with 95 billion active parameters and is accessible through QwenCloud. Qwen3.8-Max is built to answer difficult questions and complete complex deliverables from start to finish. Its coding capabilities include autonomous project creation, self-testing, issue dispatch, CI validation, pull request workflows, and long-running feedback loops. The model is also designed for real-world work across legal review, UI/UX design, restaurant operations, structural engineering, rehabilitation visualization, sports analytics, and quantitative research. Its multimodal capabilities support images, documents, videos, interface reconstruction, visual production, application recreation, and visual feedback loops. QwenCloud supports industry-standard API protocols, including OpenAI-compatible chat completions and responses APIs as well as an Anthropic-compatible interface. By combining large-scale reasoning, agentic coding, multimodal intelligence, API access, long-context workflows, and open-weight availability, Qwen3.8-Max gives teams a powerful foundation for building advanced AI systems.

GPT-5.6 Terra is OpenAI’s balanced GPT-5.6 model for users who need strong performance across everyday work, development tasks, enterprise workflows, and technical analysis. The model is part of the GPT-5.6 family alongside Sol and Luna, with Terra positioned as the middle tier for capable, cost-efficient use. Terra is described as having competitive performance to GPT-5.5 while being 2x cheaper, making it useful for teams that want advanced capability without always using the flagship model. It supports coding workflows, agentic tasks, cybersecurity-related defensive work, biology workflows, knowledge work, and tool-assisted automation. In benchmark previews, Terra appears alongside Sol and Luna in evaluations for coding, biology, ExploitBench, and ExploitGym. The model benefits from the GPT-5.6 safeguard stack, which includes model-level refusals for prohibited cyber assistance, real-time cyber and biology misuse classifiers, and account-level risk review. These safeguards are designed to preserve access to legitimate work such as code review, debugging, vulnerability research, patch development, security education, and defensive testing. GPT-5.6 Terra is planned for availability through the API, Codex, and broader OpenAI products after the limited preview period. GPT-5.6 Terra helps teams get a balanced model for high-quality AI work when they need strong reasoning and automation at a lower cost than Sol.

DeepSeek-V4-Pro is an advanced Mixture-of-Experts language model built for high-performance reasoning, coding, and large-scale AI applications. With 1.6 trillion total parameters and 49 billion activated parameters, it delivers strong capabilities while maintaining computational efficiency. The model supports a massive context window of up to one million tokens, making it ideal for handling long documents and complex workflows. Its hybrid attention architecture improves efficiency by reducing computational overhead while maintaining accuracy. Trained on more than 32 trillion tokens, DeepSeek-V4-Pro demonstrates strong performance across knowledge, reasoning, and coding benchmarks. It includes advanced training techniques such as improved optimization and enhanced signal propagation for better stability. The model offers multiple reasoning modes, allowing users to choose between faster responses or deeper analytical thinking. It is designed to support agentic workflows and complex multi-step problem solving. As an open-source model, it provides flexibility for developers and organizations to customize and deploy at scale. Overall, DeepSeek-V4-Pro delivers a balance of performance, efficiency, and scalability for demanding AI applications.

SWE-2 is a software engineering model from Cognition built for agentic coding tasks that require strong performance at lower computational and monetary cost. It is post-trained from the Kimi K3 base model and extends Cognition’s earlier SWE-1.7 training approach with a new reinforcement learning method for jointly optimizing multiple reasoning-effort settings. Medium, high, and maximum effort modes provide different tradeoffs between speed, cost, exploration, and verification depending on task complexity. The model is trained to inspect only the parts of a codebase that are likely to matter, helping it reach implementation faster and reduce unnecessary exploration. SWE-2 can generate and modify code, run tests, analyze repositories, work through terminal tasks, and verify whether implementations satisfy user requirements. Cognition also reports improvements in end-to-end test creation, regression detection, instruction following, and re-deriving conclusions when challenged. Its training process incorporates cost-aware rewards, length-weighted reward baselines, expanded reinforcement learning environments, and hardened verifiers intended to improve both efficiency and reliability. SWE-2 is positioned as a cost-efficient alternative to larger frontier coding models while remaining competitive on software engineering benchmarks such as FrontierCode, DeepSWE, and Terminal-Bench. The model is available in Devin Desktop and Devin CLI and is being introduced to additional Cognition products including Devin Web and Fusion.

Laguna S 2.1 is an advanced open weight coding model that emphasizes long-term project completion and efficient reasoning capabilities. Featuring a 118-billion-parameter Mixture-of-Experts architecture, it activates 8 billion parameters for each token and accommodates a context window of up to one million tokens in both thinking and non-thinking modes. The model’s streamlined active size allows it to perform intricate tasks on local machines while still competing favorably against significantly larger models across various benchmarks, including terminal usage, software engineering, codebase question answering, and tool utilization. Designed for resilience, Laguna S 2.1 excels in tackling challenging assignments with enhanced persistence, meticulous verification, and a readiness to backtrack rather than prematurely claim success. In practical applications, it has successfully created and validated a browser rendering engine from scratch, optimized an agent harness for improved execution speed and reduced memory usage, and conducted extensive mathematical research using the available tools within its environment, demonstrating its versatility and effectiveness. This combination of features positions Laguna S 2.1 as a powerful tool for developers seeking innovative solutions.

Kimi K2.7 Code is a Moonshot AI coding model built to help developers handle software engineering, code generation, debugging, and agent-based development workflows. It focuses on long-horizon coding tasks, where an AI assistant needs to understand goals, work across many files, and complete multi-step development work. The model builds on the Kimi K2.6 architecture and is described as improving agentic capabilities while reducing thinking-token usage by about 30% compared with K2.6. Kimi K2.7 Code offers a 256K context window, which helps developers work with larger repositories, longer prompts, and more detailed project instructions. It can be accessed through Kimi Code, Moonshot’s API platform, and third-party model providers such as Together AI. The model also supports OpenAI- and Anthropic-compatible APIs, making it easier for teams to test it as a replacement or addition to existing coding assistant workflows. Developers who want to self-host or experiment with the model can access it through Hugging Face, where deployment guidance references vLLM, SGLang, and KTransformers. Kimi K2.7 Code is especially relevant for teams interested in open-source coding agents, long-context software tasks, and tool-integrated development. While some third-party commentary notes that benchmark claims should be reviewed carefully, the model is positioned as a strong option for developers seeking flexible, agentic coding support.

Muse Spark 1.1 is Meta’s upgraded multimodal reasoning model designed to support advanced agentic workflows, coding tasks, computer use, and complex tool orchestration. Developed by Meta Superintelligence Labs, it builds on Muse Spark with major gains in planning, tool use, long-context reasoning, multimodal perception, and real-world task execution. The model can work across external apps and services, native tools, MCP servers, custom skills, browsers, scripts, images, video, PDFs, and audio inputs. Muse Spark 1.1 can act as a main agent by gathering context, creating a plan, and delegating work to parallel subagents, or operate as a subagent that follows instructions and escalates when needed. Its 1 million token context window allows it to retain earlier actions, retrieve information from long workflows, and compact context while preserving critical details. The model is also trained for computer-use tasks, deciding when to automate with scripts and when to interact directly with an interface. In coding workflows, Muse Spark 1.1 can diagnose bugs, implement features, migrate large codebases, generate web applications, take screenshots, identify UI issues, and validate fixes. Its multimodal strengths include visual-to-code generation, detailed image and video captioning, grounded perception, and workflows where seeing, reasoning, and acting happen together. Available through the Meta Model API public preview and in Thinking mode inside Meta AI, Muse Spark 1.1 gives developers and users a more capable foundation for building agents, automations, coding assistants, and multimodal productivity tools.

Inkling is Thinking Machines’ open-weights foundation model built for customization, multimodal reasoning, and agentic AI workflows. The model uses a Mixture-of-Experts architecture with 975 billion total parameters and 41 billion active parameters, making it large in capacity while activating only a subset of experts per token. Inkling supports up to a 1 million token context window and was pretrained on 45 trillion tokens spanning text, images, audio, and video. It is designed as a broad generalist model with strengths across coding, reasoning, instruction following, factuality, tool use, vision, audio understanding, forecasting, and safety. Developers can tune its thinking effort to trade off latency, cost, and performance, which is useful for production systems that need efficient reasoning at scale. Inkling can be fine-tuned on Tinker, tested in the Inkling Playground, and deployed through partners such as TogetherAI, Fireworks, Modal, Databricks, Baseten, vLLM, SGLang, llama.cpp, and Hugging Face transformers. The model can generate applications, operate tools, create styled artifacts, reason over visual and audio inputs, and support long refinement loops for collaborative work. Thinking Machines also previewed Inkling-Small, a lighter Mixture-of-Experts model with 276 billion total parameters and 12 billion active parameters for lower-cost and lower-latency workloads. By combining open weights, multimodal training, agentic capabilities, efficient reasoning, and fine-tuning support, Inkling gives builders a flexible AI foundation for specialized products and workflows.

Gemini 3.5 Flash Cyber is a dedicated model designed specifically for cybersecurity, built upon Gemini 3.5 Flash, and refined to efficiently discover, validate, and resolve vulnerabilities at scale. Its primary objective is to support defensive security operations by enabling organizations to quickly pinpoint critical vulnerabilities and produce dependable patches before they can be exploited. The remarkable blend of performance and efficiency offered by Flash provides an excellent basis for code scanning, assessing security issues, confirming the authenticity of findings, and suggesting precise remediation strategies within extensive software environments. In the CodeMender framework, numerous Gemini 3.5 Flash Cyber agents collaborate seamlessly, merging their insights into a comprehensive report that enhances the system's ability to analyze vulnerabilities from various perspectives and elevate the overall quality of the findings. This collaborative agent framework ensures exceptional performance on CyberGym, which serves as a benchmark for assessing cybersecurity effectiveness, while also fostering continuous improvement in vulnerability management practices. Ultimately, the capabilities of Gemini 3.5 Flash Cyber not only streamline security workflows but also strengthen an organization's resilience against potential threats.

Muse Spark 1.2 is Meta’s newest coding-focused model, released alongside Muse Code as part of Meta’s AI developer platform. The model improves on Muse Spark 1.1 with stronger code generation, complex debugging, codebase understanding, and full developer workflow performance. Muse Spark 1.2 powers Muse Code, a terminal coding agent that can plan changes, write code, validate results, and coordinate persistent background subagents. The model was co-trained with Muse Code so it performs well inside the agentic coding runtime and tool environment. Its training included scaled coding compute, broader training environment diversity, rejection-sampled harness trajectories, recipe optimizations, and Muse Code toolset integration. Muse Spark 1.2 is designed for long-horizon coding tasks such as whole-repository generation, large end-to-end projects, auto-research, and extended optimization work. It uses planning to sequence work, goal conditioning to stay aligned with the user’s objective, and context compaction to preserve useful knowledge over long sessions. The model also benefits from a self-improvement loop where Muse Spark 1.1 generated challenging coding environments and instruction-following templates for training. By combining coding specialization, agentic workflow support, long-horizon training, subagent compatibility, and Meta Model API availability, Muse Spark 1.2 helps developers build, debug, and optimize software more effectively.

Nemotron 3 Nano is a small yet powerful large language model from NVIDIA's Nemotron 3 series, specifically crafted for effective agentic reasoning, interactive dialogue, and programming assignments. Its innovative Mixture-of-Experts Mamba-Transformer framework selectively activates a limited set of parameters for each token, ensuring rapid inference times without sacrificing accuracy or reasoning capabilities. With roughly 31.6 billion parameters in total, including about 3.2 billion active ones (or 3.6 billion when factoring in embeddings), it surpasses the performance of the previous Nemotron 2 Nano model while requiring less computational effort for each forward pass. The model is equipped to manage long-context processing of up to one million tokens, which allows it to efficiently process extensive documents, complex workflows, and detailed reasoning sequences in a single cycle. Moreover, it is engineered for high-throughput, real-time performance, making it particularly adept at handling multi-turn dialogues, invoking tools, and executing agent-based workflows that involve intricate planning and reasoning tasks. This versatility positions Nemotron 3 Nano as a leading choice for applications requiring advanced cognitive capabilities.

MiniMax M3 is a frontier open-weight AI model built for coding, agentic work, multimodal understanding, and ultra-long-context tasks. The model supports up to a 1 million token context window, allowing it to work across large codebases, long documents, logs, project histories, and complex task environments. MiniMax M3 introduces MiniMax Sparse Attention, a sparse attention architecture designed to make long-context processing more efficient. The model is natively multimodal, with training that supports deeper semantic fusion across text, image, and video inputs. It is designed to support software engineering tasks, repository analysis, terminal-style work, browser-style retrieval, tool use, and autonomous workflows. MiniMax M3 has a mixture-of-experts architecture with hundreds of billions of total parameters and a smaller activated parameter count for more efficient inference. Developers can use it for AI coding assistants, workflow automation, research agents, document analysis, visual reasoning, and enterprise AI systems. Its long-context capability makes it especially useful when tasks require many files, references, instructions, or interaction histories to stay available at once. MiniMax M3 helps teams build more capable AI agents that can understand larger problems, work across multiple modalities, and execute complex tasks with stronger context awareness.

DeepSeek-V3.2 is a highly optimized large language model engineered to balance top-tier reasoning performance with significant computational efficiency. It builds on DeepSeek's innovations by introducing DeepSeek Sparse Attention (DSA), a custom attention algorithm that reduces complexity and excels in long-context environments. The model is trained using a sophisticated reinforcement learning approach that scales post-training compute, enabling it to perform on par with GPT-5 and match the reasoning skill of Gemini-3.0-Pro. Its Speciale variant overachieves in demanding reasoning benchmarks and does not include tool-calling capabilities, making it ideal for deep problem-solving tasks. DeepSeek-V3.2 is also trained using an agentic synthesis pipeline that creates high-quality, multi-step interactive data to improve decision-making, compliance, and tool-integration skills. It introduces a new chat template design featuring explicit thinking sections, improved tool-calling syntax, and a dedicated developer role used strictly for search-agent workflows. Users can encode messages using provided Python utilities that convert OpenAI-style chat messages into the expected DeepSeek format. Fully open-source under the MIT license, DeepSeek-V3.2 is a flexible, cutting-edge model for researchers, developers, and enterprise AI teams.

Aya Expanse revolutionizes the field of multilingual AI by serving as a research model that adeptly handles 101 languages, utilizing cutting-edge instruction tuning and cross-lingual transfer methods. The model's unique approach merges a carefully selected open source dataset with efficient pretraining processes, allowing it to deliver exceptional results for both low- and high-resource languages. This innovation not only enhances performance but also successfully lowers infrastructure expenses by up to 30%, establishing a new standard for scalable and inclusive language modeling in the industry. As a result, Aya Expanse is poised to make a significant impact on the future of AI language processing.

DeepSeek-V4-Flash is an optimized Mixture-of-Experts language model built for efficient large-scale AI workloads and fast inference. With 284 billion total parameters and 13 billion activated parameters, it delivers strong performance while maintaining lower computational demands compared to larger models. The model supports a massive context length of up to one million tokens, making it suitable for handling long-form content and multi-step workflows. Its hybrid attention mechanism improves efficiency by minimizing resource consumption while preserving accuracy. Trained on a dataset exceeding 32 trillion tokens, DeepSeek-V4-Flash performs well across reasoning, coding, and knowledge benchmarks. It offers flexible reasoning modes, enabling users to switch between quick responses and more detailed analytical outputs. The architecture is designed to support agentic workflows and scalable deployment environments. As an open-source model, it provides flexibility for customization and integration. Overall, DeepSeek-V4-Flash is a cost-effective and high-performance solution for modern AI applications.

DeepSeek-V4 is an advanced open-source large language model engineered for efficient long-context processing and high-level reasoning tasks. Supporting a massive one million token context window, it enables developers to build applications that handle extensive data and complex workflows without fragmentation. The model is available in two versions: V4-Pro for maximum reasoning power and V4-Flash for faster, cost-efficient performance. DeepSeek-V4-Pro delivers top-tier results in coding, mathematics, and knowledge benchmarks, rivaling leading proprietary models. Its architecture incorporates innovative attention techniques that significantly improve efficiency while maintaining strong performance. The model is optimized for agent-based workflows, allowing seamless integration with tools and automation systems. It also supports dual reasoning modes, enabling users to switch between quick responses and deeper analytical outputs. DeepSeek-V4 is fully open-source, providing flexibility for customization and deployment across various environments. Overall, it offers a powerful and scalable solution for modern AI development.

DeepSeek-V4.1-Flash-1017 is an open-weight multimodal AI model designed for complex reasoning, autonomous agents, software development, and visual analysis. The model is an updated version of DeepSeek-V4.1-Flash, with improvements in benchmark performance, execution speed, and output efficiency. It supports text and image processing, enabling applications that combine natural language understanding with visual interpretation and reasoning. Its one-million-token context window allows developers to analyze large documents, extensive software repositories, and lengthy interactions while maintaining contextual information. Coding capabilities include generating software, debugging existing applications, modifying codebases, and completing multistep engineering assignments. The model also supports end-to-end project generation, including interactive games, 3D scenes, and other software applications. Agentic functionality includes external tool calling, configurable reasoning effort, multi-turn conversations, and workflows involving automated research and task execution. DeepSeek provides prompt encoding libraries, inference examples, and compatibility with deployment frameworks such as Hugging Face Transformers, vLLM, and SGLang. DeepSeek-V4.1-Flash-1017 is distributed under the MIT License and can be accessed through DeepSeek's website, mobile application, and API or deployed using its publicly available model weights.

DeepSeek-V4.1-Flash is a highly efficient and adaptable AI model tailored for complex tasks in coding, creativity, agentic functions, and spatial reasoning. Building on its predecessor, DeepSeek-V4-Flash, this version prioritizes rapid output generation while ensuring robust performance on intricate challenges, achieving over 400 tokens per second with peak performance reaching approximately 427 tokens per second in tests. The model is equipped to handle sophisticated programming tasks, craft immersive 3D environments, develop voxel-based creations, and analyze spatially intricate scenes and simulations. Noteworthy demonstrations showcase its versatility through Minecraft-inspired worlds, traditional Chinese gardens, racing tracks, dungeon exploration, exploded camera perspectives, and other scenarios that require a blend of coding skills and spatial comprehension. Its advanced capabilities position it as an ideal choice for quick prototyping, game design, 3D modeling, architecture, academic research, and various other technical or artistic endeavors where speed in iteration is crucial. Additionally, the model's innovative features allow it to adapt to diverse project requirements, further enhancing its usefulness across multiple domains.

DiffusionGemma is an innovative open model that investigates text diffusion, representing a remarkably rapid method for generating text. Released under the Apache 2.0 license, this 26 billion parameter Mixture of Experts (MoE) model advances beyond the usual sequential token generation typical of autoregressive models. Instead, it produces entire blocks of text at once, achieving text generation speeds that are up to four times faster on GPUs. Drawing from the parameter efficiency of the Gemma 4 family and Gemini Diffusion research, DiffusionGemma incorporates a unique diffusion head that enhances generation speed significantly. It is particularly aimed at researchers and developers looking to optimize speed-sensitive, interactive local workflows, including in-line editing, swift iterations, and non-linear narrative forms. By reallocating the decode bottleneck from memory bandwidth to computational power, it can produce over 1,000 tokens per second on a single NVIDIA H100 and more than 700 tokens per second on an NVIDIA GeForce RTX 5090. This breakthrough allows for a new level of efficiency in text generation that could reshape various applications in natural language processing.

Command A+ represents Cohere’s most advanced and rapid language model to date, serving as a robust open-source tool tailored for intricate reasoning, diverse multimodal and multilingual tasks, and seamless private deployment. With its architecture as a sparse mixture-of-experts, it boasts a remarkable 218 billion total parameters, of which 25 billion are actively utilized, ensuring high-performance agentic workflows while minimizing computational demands. This model consolidates features from the entire Command series into a single scalable solution, accommodating text, images, reasoning, and tool utilization with an impressive 128K input context, a maximum generation of 64K, and compatibility with 48 different languages. It has been meticulously optimized to enhance reasoning capabilities, agentic workflows, retrieval-augmented generation (RAG), multilingual applications, and the processing of multimodal documents, while also supporting vLLM and Transformers technology. When compared to its predecessors in the Command A lineup, it significantly boosts enterprise performance across various domains, including multimodal comprehension, data retrieval, extended tasks, sophisticated reasoning, programming, translation, and thorough document analysis. The advancements in this model underline its potential to transform how enterprises approach complex language and data processing challenges.

Kimi K2.5 is a powerful multimodal AI model built to handle complex reasoning, coding, and visual understanding at scale. It supports both text and image or video inputs, enabling developers to build applications that go beyond traditional language-only models. As Kimi’s most advanced model to date, it delivers open-source state-of-the-art performance across agent tasks, software development, and general intelligence benchmarks. The model supports an ultra-long 256K context window, making it ideal for large codebases, long documents, and multi-turn conversations. Kimi K2.5 includes a long-thinking mode that excels at logical reasoning, mathematics, and structured problem solving. It integrates seamlessly with existing workflows through full compatibility with the OpenAI SDK and API format. Developers can use Kimi K2.5 for chat, tool calling, file-based Q&A, and multimodal analysis. Built-in support for streaming, partial mode, and web search expands its flexibility. With predictable pricing and enterprise-ready capabilities, Kimi K2.5 is designed for scalable AI development.

Grok 4.8 is an upcoming large language model from xAI designed to continue the company’s push toward more capable coding, reasoning, knowledge work, and autonomous AI agents. Elon Musk has stated that Grok 4.8 uses approximately 2.5 trillion parameters, making it larger than the 2.1-trillion-parameter Grok 4.7 model previously discussed. The model is also being trained on a new C++ software stack that xAI expects to use for its next generation of large-scale training runs. Initial model training is expected to finish before reinforcement learning and additional post-training work begin, meaning the final production model is not yet available. Based on the current Grok generation, Grok 4.8 is likely to emphasize software engineering, agentic tool use, professional knowledge work, image understanding, and complex multi-step reasoning. Grok 4.7 currently supports configurable reasoning levels and a 500,000-token context window, providing a baseline for the capabilities xAI is developing further. Grok 4.8 may also become an important model for products such as Grok Build and Grok Bot, where stronger reasoning and tool coordination can support longer autonomous workflows. xAI has not published official Grok 4.8 benchmarks, pricing, context limits, API details, or an exact release date. Grok 4.8 is expected to serve developers, engineering teams, researchers, enterprises, and AI agent builders seeking frontier-level performance across technical and professional tasks.

Laguna M.1 stands out as Poolside's most proficient model for agentic coding, meticulously developed in-house specifically for enhancing software development workflows. This model features a total of 225 billion parameters, utilizing a Mixture of Experts architecture with 23 billion activated parameters, and has been trained entirely within the organization on a dataset consisting of 30 trillion tokens, leveraging the power of 6,144 interconnected NVIDIA H200 GPUs. Poolside undertook the task of training Laguna M.1 from the ground up, employing its proprietary data, dedicated training codebase, and an asynchronous on-policy reinforcement learning approach within its agent framework, all tailored for agentic coding applications. The design of the model ensures optimal performance within Poolside's coding agent, enabling it to effectively reason through software tasks, interact with various tools, edit code, execute tests, and facilitate extended autonomous development sessions. Specifically crafted for developers and teams tackling intricate coding challenges, Laguna M.1 offers enhanced capabilities in reasoning, architectural comprehension, terminal operations, and multi-step execution, surpassing what lighter models can achieve. Ultimately, its robust feature set positions it as an essential asset for those engaged in demanding software projects.

Kimi K2.6 is an advanced agentic AI model created by Moonshot AI, aiming to enhance practical implementation, programming, and complex reasoning compared to its predecessors, K2 and K2.5. This model is based on a Mixture-of-Experts framework and the multimodal, agent-centric principles of the Kimi series, merging language comprehension, coding capabilities, and tool utilization into one cohesive system that can plan and execute intricate workflows. It features enhanced reasoning skills and significantly better agent planning, enabling it to deconstruct tasks, synchronize various tools, and tackle multi-file or multi-step challenges with increased precision and effectiveness. Additionally, it provides robust tool-calling capabilities with a high degree of reliability, facilitating seamless integration with external platforms like web searches or APIs, and incorporates built-in validation systems to guarantee the accuracy of execution formats. Notably, Kimi K2.6 represents a significant leap forward in the realm of AI, setting new standards for the complexity and reliability of automated tasks.

North Mini Code marks the debut of Cohere’s agentic coding model tailored for developers and serves as the first entry in its next generation of robust models. This compact and efficient open-source solution is specifically crafted for the independent developer community, ensuring remarkable software development capabilities without the need for high-end hardware. Featuring a mixture-of-experts architecture, it comprises a total of 30 billion parameters, with 3 billion of those being active, thereby providing developers with powerful agentic coding functionalities in a streamlined package. The model is finely tuned for various tasks, including code generation, agentic software engineering, and terminal operations, boasting an impressive 256K context length and a maximum generation capacity of 64K. It is designed with real-world developer practices in mind, enabling tasks such as understanding and managing sub-agents, mapping out system architectures, conducting code reviews, and assisting coding agents in navigating intricate software challenges. The integration of these capabilities empowers developers to enhance their productivity and efficiency significantly in software development projects.

Laguna XS.2 represents Poolside’s innovative open-weight coding model, distinguished as the lightest and quickest member of the Laguna series. This model features a total of 33 billion parameters in a Mixture of Experts setup, with 3 billion parameters activated, and has been meticulously trained in-house using 30 trillion tokens. As the latest generation model accessible to the public, it embodies a second-generation architecture and marks Poolside’s inaugural open-weight offering, drawing from insights gained during the training of Laguna M.1 with synthetic data and reinforcement learning techniques. Specifically designed to enhance agentic coding workflows, Laguna XS.2 excels in coding, acting, and rapidly iterating, particularly within Poolside’s coding agent environment. This model is particularly advantageous for developers and teams seeking a lightweight, efficient coding solution rather than a more cumbersome frontier system. Released under the permissive Apache 2.0 license, it empowers the community to assess, fine-tune, quantize, and build upon its weights, fostering a collaborative development atmosphere. In essence, Laguna XS.2 not only provides a robust platform for agentic coding but also encourages innovation and experimentation among its users.

Muse Spark is Meta’s first model in the Muse family, designed as a natively multimodal AI system focused on advanced reasoning and real-world applications. It combines text, visual understanding, and tool usage to provide more interactive and context-aware responses. The model introduces capabilities like visual chain-of-thought reasoning and multi-agent orchestration for complex problem-solving. Its Contemplating mode allows multiple AI agents to work in parallel, improving accuracy on challenging tasks. Muse Spark performs strongly across domains such as STEM reasoning, health insights, and multimodal perception. It can analyze images, generate interactive outputs, and assist with tasks like troubleshooting or educational content. The model is trained using improved pretraining, reinforcement learning, and efficient test-time reasoning techniques. It is designed to scale efficiently while delivering high performance with optimized compute usage. Safety measures include strong refusal behavior and alignment safeguards across high-risk domains. Overall, Muse Spark is a foundational step toward building personalized, highly capable AI systems.