Alternatives to Merrymake

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.

Claude Opus 5 is Anthropic’s new state-of-the-art Opus model for coding, knowledge work, automation, science, and everyday AI assistance. The model is designed to provide near-frontier intelligence at a lower cost than Claude Fable 5 while keeping the same base pricing as Opus 4.8. Claude Opus 5 performs strongly on software engineering benchmarks, business task automation, computer use, novel problem solving, visual generation, and scientific research tasks. Users can adjust effort settings to trade off intelligence, speed, and token usage depending on the task. The model is also better at verifying its work, iterating carefully, building test harnesses, debugging root causes, and solving multi-step engineering problems. Anthropic highlights improvements in life sciences, including structural biology, organic chemistry, bioinformatics, and protein-related tasks. Claude Opus 5 includes alignment and safety safeguards designed to support beneficial work while blocking higher-risk cybersecurity and biology misuse. It is available through Claude.ai, Claude Max, Claude Pro, Claude Code, Claude Cowork, and the Claude API under the model name claude-opus-5. By combining stronger reasoning, coding ability, scientific capability, configurable effort, Fast mode, and enterprise-ready deployment options, Claude Opus 5 gives users a powerful model for demanding daily work.

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.

Claude Mythos 5 is a frontier AI model from Anthropic created for highly trusted users working on advanced cybersecurity, infrastructure protection, and scientific research. It is based on the same core model as Claude Fable 5, but certain safeguards are lifted for approved partners operating under restricted access programs. The model offers exceptional performance across software engineering, cybersecurity analysis, autonomous development workflows, scientific reasoning, visual understanding, and long-context tasks. In cybersecurity, Claude Mythos 5 is positioned for cyberdefenders and critical infrastructure providers who need advanced AI support for securing complex systems. In life sciences, the model has demonstrated strong capabilities in drug design, protein research, molecular biology, and genomics. Claude Mythos 5 can perform long-running research and technical workflows with minimal high-level human input. Anthropic designed the model for controlled deployment because its advanced capabilities could create misuse risks if broadly available without safeguards. Access is initially limited to Project Glasswing partners, with broader trusted access programs planned for cybersecurity and select biology researchers. Claude Mythos 5 helps approved organizations apply powerful AI to high-impact technical and scientific challenges while operating within a stricter governance model.

Sakana Fugu Ultra is a performance-optimized multi-agent AI model designed for hard technical, research, security, and analytical workloads. It coordinates a deeper pool of expert agents than the standard Fugu model, allowing it to focus on maximum answer quality for complex tasks. The model is available through the same OpenAI-compatible API as Sakana Fugu, making it easier to integrate into existing tools, developer workflows, and AI applications. Fugu Ultra is especially useful for coding, advanced code review, Kaggle competitions, paper reproduction, cybersecurity assessments, literature reviews, patent research, and long-running autonomous workflows. Instead of requiring users to choose individual models or define agent roles, Fugu Ultra dynamically assembles and coordinates the agents that are best suited for each task. Its approach is grounded in learned model orchestration research, including TRINITY and the Conductor, which explore how multiple AI systems can collaborate more effectively. Organizations can also control which providers or models participate in the agent pool to support privacy, compliance, and internal policy requirements. Fugu Ultra is positioned for high-value tasks where deeper analysis, stronger reasoning, and better reliability matter more than speed alone. Sakana Fugu Ultra gives developers, researchers, and enterprises a way to use frontier-level multi-agent intelligence through one managed endpoint.

Gemini 3.5 Flash is Google’s high-performance multimodal AI model built to deliver frontier-level intelligence, fast execution speeds, and advanced agentic capabilities for coding, automation, and enterprise workflows. As the first release in the Gemini 3.5 series, the model is designed to help developers, businesses, and users execute complex long-horizon tasks through AI-powered reasoning, workflow orchestration, and intelligent automation. Gemini 3.5 Flash combines powerful coding performance, multimodal understanding, and real-time responsiveness while outperforming earlier Gemini models and competing frontier AI systems across several coding and reasoning benchmarks. The model is optimized for agentic workflows, allowing it to plan, execute, and manage multi-step tasks such as software development, infrastructure management, document preparation, and business process automation through the updated Antigravity harness. Gemini 3.5 Flash can also deploy collaborative subagents that work together under supervision to complete demanding workflows more efficiently and at lower operational cost. Beyond coding and automation, the platform generates richer graphics, dynamic web interfaces, interactive animations, and advanced multimodal experiences that support developers and enterprise users building AI-driven applications. Google has integrated Gemini 3.5 Flash across the Gemini app, AI Mode in Google Search, Google AI Studio, Android Studio, Gemini Enterprise Agent Platform, and enterprise AI services to expand access to advanced AI capabilities globally. The model also powers Gemini Spark, Google’s new personal AI agent designed to operate continuously and assist users with digital life management and automated task execution.

FLUX.2 advances the FLUX model family with major improvements in realism, prompt adherence, and world knowledge, enabling it to produce coherent lighting, spatial logic, and accurate material properties. It offers multi-reference generation with support for up to 10 images, allowing creators to maintain continuity across characters, products, and environments. The model reliably handles complex text, detailed typography, and branding requirements, making it suitable for marketing, design, and enterprise workflows. Editing capabilities reach resolutions up to 4 megapixels, preserving fine structure and stylistic fidelity. FLUX.2 is built on a latent flow matching architecture, combining a Mistral-3 based vision-language model with a rectified-flow transformer to unify generation and editing. Its variants—FLUX.2 [pro], FLUX.2 [flex], FLUX.2 [dev], and the upcoming FLUX.2 [klein]—offer a full spectrum of performance and control for teams of all sizes. Developers can self-host open weights, integrate via API, or tune generation parameters for full-stack customization. In every configuration, FLUX.2 is designed to radically improve productivity while lowering the cost of high-quality image creation.

Fugu Cyber is an advanced orchestration model designed specifically for contemporary cyber defense, operating as a unified entity through a single API endpoint while adeptly managing multiple specialized agents to tackle intricate security challenges. This innovative model does not rely on a single provider and is tailored for two main defense operations: assessing complex codebases to identify genuine vulnerabilities and converting raw cyber threat intelligence into actionable detection rules. Its performance on CyberGym, which tests vulnerability analysis and validation, resulted in an impressive success rate of 86.9%, whereas on CTI-REALM, which evaluates the generation of detection rules from threat intelligence reports, it achieved a score of 72.1%. These results position Fugu Cyber among the top-tier models focused on cybersecurity innovations. Rather than functioning as an isolated tool, Fugu Cyber is designed to serve as the cognitive engine within larger security infrastructures, enhancing overall defense capabilities against evolving cyber threats. This integration allows for a more holistic approach to cyber defense, enabling organizations to respond more effectively to potential attacks.

Zyphra Cloud serves as a comprehensive platform aimed at fostering open superintelligence, translating cutting-edge advancements from Zyphra Research into practical applications for developers, businesses, and leading AI hyperscalers. Tailored for sophisticated AI solutions, it emphasizes the development of long-term agents by integrating agent infrastructure, inference, agent environments, and computational resources into a cohesive system designed for the construction and deployment of open, sovereign AI at a grand scale. Among its features, Zyphra Cloud boasts MAIA, a versatile open superagent crafted for teamwork: a cohesive multimodal framework that harmonizes knowledge sharing, communication, and task execution across various tools and workflows. Designed with multiplayer functionality, MAIA ensures a shared context, maintains persistent memory, and allows for synchronized operations among users and tools, facilitating interactions through language, audio, and visual inputs within a singular, unified reasoning framework. The platform’s initial offering, Zyphra Inference, is specifically engineered to cater to the demands of long-horizon agentic workloads, ensuring efficiency and performance. Furthermore, the integration of these components aims to empower users to innovate and enhance their AI capabilities seamlessly.

SWE-1.7 is Cognition’s most capable software engineering model, built to push frontier coding performance while reducing the cost of high-quality agentic rollouts. The model is designed for real-world software development tasks that require extended reasoning, codebase understanding, terminal use, debugging, feature work, migrations, and careful validation. It was trained from a Kimi K2.7 base and improved through Cognition’s reinforcement learning pipeline, including more stable training, stronger infrastructure, better data curation, and long-horizon task techniques. SWE-1.7 is especially optimized for asynchronous software engineering, where an agent needs to work through large projects over longer sessions instead of simply answering short prompts. Its self-compaction capabilities allow the model to summarize its working state and resume from that summary, helping it operate beyond the raw context window on multi-hour tasks. The model is also trained to balance task success with efficiency, using concise reasoning when possible while preserving deeper exploration for harder problems. SWE-1.7 tends to investigate codebases more thoroughly than its base model, reading files, running searches, probing edge cases, and experimenting before making changes. It is available in Devin through web, desktop, and CLI interfaces, with Cerebras serving support at 1000 TPS. SWE-1.7 gives developers and engineering teams a high-performance coding model for complex software projects at a more practical cost.

Step 3.5 Flash is a cutting-edge open-source foundational language model designed for advanced reasoning and agent-like capabilities, optimized for efficiency; it utilizes a sparse Mixture of Experts (MoE) architecture that activates only approximately 11 billion of its nearly 196 billion parameters per token, ensuring high-density intelligence and quick responsiveness. The model features a 3-way Multi-Token Prediction (MTP-3) mechanism that allows it to generate hundreds of tokens per second, facilitating complex multi-step reasoning and task execution while efficiently managing long contexts through a hybrid sliding window attention method that minimizes computational demands across extensive datasets or codebases. Its performance on reasoning, coding, and agentic tasks is formidable, often matching or surpassing that of much larger proprietary models, and it incorporates a scalable reinforcement learning system that enables continuous self-enhancement. Moreover, this innovative approach positions Step 3.5 Flash as a significant player in the field of AI language models, showcasing its potential to revolutionize various applications.

Mistral Compute is a specialized AI infrastructure platform that provides a comprehensive, private stack including GPUs, orchestration, APIs, products, and services, available in various configurations from bare-metal servers to fully managed PaaS solutions. Its mission is to broaden access to advanced AI technologies beyond just a few providers, enabling governments, businesses, and research organizations to design, control, and enhance their complete AI landscape while training and running diverse workloads on an extensive array of NVIDIA-powered GPUs, all backed by reference architectures crafted by experts in high-performance computing. This platform caters to specific regional and sectoral needs, such as defense technology, pharmaceutical research, and financial services, and incorporates four years of operational insights along with a commitment to sustainability through decarbonized energy sources, ensuring adherence to strict European data-sovereignty laws. Additionally, Mistral Compute’s design not only prioritizes performance but also fosters innovation by allowing users to scale and customize their AI applications as their requirements evolve.

Kimi K2 represents a cutting-edge series of open-source large language models utilizing a mixture-of-experts (MoE) architecture, with a staggering 1 trillion parameters in total and 32 billion activated parameters tailored for optimized task execution. Utilizing the Muon optimizer, it has been trained on a substantial dataset of over 15.5 trillion tokens, with its performance enhanced by MuonClip’s attention-logit clamping mechanism, resulting in remarkable capabilities in areas such as advanced knowledge comprehension, logical reasoning, mathematics, programming, and various agentic operations. Moonshot AI offers two distinct versions: Kimi-K2-Base, designed for research-level fine-tuning, and Kimi-K2-Instruct, which is pre-trained for immediate applications in chat and tool interactions, facilitating both customized development and seamless integration of agentic features. Comparative benchmarks indicate that Kimi K2 surpasses other leading open-source models and competes effectively with top proprietary systems, particularly excelling in coding and intricate task analysis. Furthermore, it boasts a generous context length of 128 K tokens, compatibility with tool-calling APIs, and support for industry-standard inference engines, making it a versatile option for various applications. The innovative design and features of Kimi K2 position it as a significant advancement in the field of artificial intelligence language processing.

Aristotle represents a groundbreaking advancement as the inaugural AI model constructed entirely as a Mathematical Superintelligence (MSI), aimed at providing solutions that are mathematically verified to intricate quantitative challenges without any instances of hallucination. When it receives inquiries in natural language related to mathematics, it translates these into Lean 4 formalism, solves them through rigorously verified proofs, and subsequently delivers both the proof alongside a natural language interpretation. In contrast to traditional language models that depend on probabilistic methods, the MSI framework of Aristotle eliminates uncertainty by employing demonstrable logic and openly identifying any errors or discrepancies. This innovative AI can be accessed via a web interface and developer API, allowing researchers to incorporate its precise reasoning capabilities into various domains, including theoretical physics, engineering, and computer science. Its design not only streamlines problem-solving but also enhances the reliability of results across multiple disciplines.

Qwen3.6-Max-Preview represents an advanced frontier language model aimed at enhancing intelligence, following instructions, and improving real-world agent functionalities within the Qwen ecosystem. This preview builds upon the Qwen3 series, showcasing enhanced world knowledge, refined alignment with instructions, and notable advancements in coding performance for agents, which allows the model to adeptly manage intricate, multi-step tasks and software engineering processes. It is meticulously designed for scenarios requiring advanced reasoning and execution, where the model goes beyond merely generating responses to actively interacting with tools, processing lengthy contexts, and facilitating structured problem-solving in various fields such as coding, research, and enterprise operations. The architecture continues to embody the Qwen commitment to developing large-scale, high-efficiency models that can effectively manage extensive context windows while providing reliable performance across multilingual and knowledge-intensive projects. Moreover, its capabilities promise to significantly enhance productivity and innovation in diverse applications.

STACKIT is a cloud computing platform based in Europe that aims to offer scalable, secure, and data-sovereign infrastructure tailored for businesses, public entities, and regulated sectors. It provides a comprehensive suite of cloud services enabling organizations to operate applications, manage data, and develop digital solutions through infrastructure and platform tools located in European data centers. The offerings encompass infrastructure-as-a-service elements, including virtual machines, storage options, and networking capabilities, alongside platform services like managed databases, container environments, and application development frameworks. Emphasizing digital sovereignty, STACKIT ensures that data handling, processing, and operational management remain within the confines of the European Union and adhere to European regulations, thus assisting organizations in complying with stringent data protection mandates such as GDPR. In addition to these features, STACKIT also prioritizes user privacy, ensuring that clients can trust their data is managed securely and in accordance with local laws.

Mistral Large 3 pushes open-source AI into frontier territory with a massive sparse MoE architecture that activates 41B parameters per token while maintaining a highly efficient 675B total parameter design. It sets a new performance standard by combining long-context reasoning, multilingual fluency across 40+ languages, and robust multimodal comprehension within a single unified model. Trained end-to-end on thousands of NVIDIA H200 GPUs, it reaches parity with top closed-source instruction models while remaining fully accessible under the Apache 2.0 license. Developers benefit from optimized deployments through partnerships with NVIDIA, Red Hat, and vLLM, enabling smooth inference on A100, H100, and Blackwell-class systems. The model ships in both base and instruct variants, with a reasoning-enhanced version on the way for even deeper analytical capabilities. Beyond general intelligence, Mistral Large 3 is engineered for enterprise customization, allowing organizations to refine the model on internal datasets or domain-specific tasks. Its efficient token generation and powerful multimodal stack make it ideal for coding, document analysis, knowledge workflows, agentic systems, and multilingual communications. With Mistral Large 3, organizations can finally deploy frontier-class intelligence with full transparency, flexibility, and control.

Cognition has unveiled SWE-1.5, the newest agent-model specifically designed for software engineering, featuring an expansive "frontier-size" architecture composed of hundreds of billions of parameters and an end-to-end optimization (encompassing the model, inference engine, and agent harness) that enhances both speed and intelligence. This model showcases nearly state-of-the-art coding capabilities and establishes a new standard for latency, achieving inference speeds of up to 950 tokens per second, which is approximately six times quicker than its predecessor, Haiku 4.5, and thirteen times faster than Sonnet 4.5. Trained through extensive reinforcement learning in realistic coding-agent environments that incorporate multi-turn workflows, unit tests, and quality assessments, SWE-1.5 also leverages integrated software tools and high-performance hardware, including thousands of GB200 NVL72 chips paired with a custom hypervisor infrastructure. Furthermore, its innovative architecture allows for more effective handling of complex coding tasks and improves overall productivity for software development teams. This combination of speed, efficiency, and intelligent design positions SWE-1.5 as a game changer in the realm of coding models.

Qwen3.7-Max represents the latest advancement in Qwen's proprietary models, tailored for the agent era, and serves as a robust foundation for various applications, including code writing and debugging, office workflow automation, and maintaining extended autonomous browser sessions. This model achieves top-tier coding performance, demonstrating superior capabilities in software engineering, terminal operations, GUI interactions, web browsing, and the utilization of agentic tools. By enhancing the alignment between model intelligence and real-world agent execution, Qwen3.7-Max facilitates advanced planning, long-context reasoning, dependable function invocation, and the execution of multi-step tasks within intricate workflows. Furthermore, it bolsters multimodal and document-centric tasks through Qwen Studio, which enables chatbot interactions, comprehends images and videos, generates images, processes documents, creates presentations, offers coding support, conducts in-depth research, and enables web development. This comprehensive suite of features positions Qwen3.7-Max as a leading solution for diverse operational needs in the modern digital landscape.

GPT-Rosalind is an advanced reasoning model created by OpenAI, aimed at enhancing scientific exploration in fields like biology, drug development, and translational medicine. Tailored for workflows in life sciences, it assists researchers in managing extensive literature, experimental findings, and specialized databases to formulate and test innovative concepts. By integrating a profound understanding of disciplines such as chemistry, genomics, protein engineering, and disease biology with sophisticated tool-usage capabilities, it effectively interacts with scientific databases, examines experimental results, and facilitates intricate, multi-stage reasoning tasks. Its functionalities span evidence synthesis, hypothesis formulation, literature assessment, sequence analysis, and experimental design, empowering scientists to transition more swiftly from raw data to meaningful insights. Furthermore, GPT-Rosalind revolutionizes cumbersome, time-consuming research methodologies into streamlined, AI-enhanced workflows, ultimately fostering a more productive scientific environment. This model exemplifies the fusion of artificial intelligence with scientific inquiry, paving the way for groundbreaking discoveries.

GPT-5.3-Codex is a next-generation AI agent built to expand Codex beyond code writing into full-spectrum professional execution. It unifies advanced coding intelligence with reasoning, planning, and computer-use capabilities. The model delivers faster performance while handling more complex workflows across development environments. GPT-5.3-Codex can autonomously iterate on large projects while remaining interactive and steerable. It supports tasks such as debugging, deployment, performance optimization, and system monitoring. The model demonstrates state-of-the-art results across real-world coding benchmarks. It also excels at web development, generating production-ready applications from minimal prompts. GPT-5.3-Codex understands intent more effectively, producing stronger default designs and functionality. Its agentic nature allows it to operate like a collaborative teammate. This makes it suitable for both individual developers and large teams.

LongCat-2.0 represents a significant advancement in the realm of language models, featuring a staggering 1.6 trillion parameters through a Mixture-of-Experts architecture that leverages AI ASIC superpods, with approximately 48 billion parameters engaged per token, showcasing exceptional capabilities in coding and agentic tasks. This model marks a notable improvement over its predecessors by integrating a large-scale sparse architecture with specialized post-training methods tailored for tasks in real-world software development, tool utilization, long-context reasoning, and complex agent workflows. Entirely developed and executed on AI ASIC superpods, LongCat-2.0 underwent pretraining that encompassed over 35 trillion tokens and millions of accelerator hours, exemplifying cutting-edge training methodologies on innovative hardware solutions. To enhance its performance on tasks requiring long-term context, the model incorporates LongCat Sparse Attention and is trained using hundreds of billions of tokens from 1M-context datasets, enabling it to effectively manage ultra-long context tasks and ensure robust understanding of lengthy documents. This combination of features positions LongCat-2.0 as a pioneering force in the landscape of advanced language models.

Arcee AI is an open intelligence lab based in the United States that focuses on developing advanced, open-weight AI models. It is dedicated to creating high-performance AI systems that are both scalable and accessible to developers and enterprises. The company’s Trinity model family showcases its ability to deliver frontier-level reasoning capabilities with transparent architectures. Arcee AI leverages techniques such as online reinforcement learning to enable continuous model improvement even after deployment. Its models are designed to handle complex tasks, multi-step reasoning, and agent-based workflows. The company emphasizes cost efficiency, ensuring that organizations can access powerful AI without the high expenses typically associated with cutting-edge models. Arcee AI also provides APIs, documentation, and tools that support seamless integration into various applications. Its open-weight approach allows developers to customize and deploy models according to their specific needs. The platform is built to support a wide range of industries, including research, enterprise automation, and AI development. Ultimately, Arcee AI is focused on democratizing access to advanced AI technologies while maintaining performance and scalability.

The new OpenAI API service tier, GPT-5.6 Sol Ultrafast, operates up to 14 times quicker than the Standard processing version, delivering cutting-edge intelligence to applications and workflows where every fleeting moment is crucial. Utilizing Cerebras technology, it boasts the capability to produce as many as 750 output tokens each second, enabling sophisticated reasoning to function at real-time velocities without the need for a more compact or specialized model. This service is particularly tailored for business environments where rapid responses can significantly enhance the capabilities of AI systems. It has various applications, including incident response, where it can swiftly analyze logs, code changes, traces, and engineering reports during ongoing outages; financial research and security, where it can rapidly evaluate fluctuating market signals and identify suspicious transactions; and customer support, where intricate problems can be resolved seamlessly during live conversations. In the realm of e-commerce, it excels at handling product inquiries, verifying inventory status, and customizing product recommendations to enhance user experience. By implementing this advanced service, organizations can expect improved efficiency and effectiveness in their operations.

Celeris-1 stands out as a swift and versatile language model platform, complemented by a diffusion model that achieves cutting-edge intelligence at unprecedented speeds. Unlike conventional autoregressive models that generate tokens sequentially, Celeris employs a diffusion-based inference architecture that allows for simultaneous generation, resulting in response times that can be measured in mere milliseconds. On the MMLU-Pro benchmark, Celeris-1 boasts an impressive accuracy of 75.9% while achieving a median response time of 158 milliseconds and producing an astonishing 1,664 output tokens per second, positioning it closely to leading models but operating over ten times faster. This powerful model is accessible through an API that is compatible with OpenAI, enabling developers to seamlessly integrate it into existing SDKs and applications with minimal modifications. Additionally, it supports streaming capabilities for real-time applications, allowing for response times as low as 24 milliseconds without any buffering or delays, making it an ideal choice for interactive use cases. Overall, Celeris-1 represents a significant advancement in the efficiency and performance of language models.

The latest advancement, GPT-4 with vision (GPT-4V), allows users to direct GPT-4 to examine image inputs that they provide, marking a significant step in expanding its functionalities. Many in the field see the integration of various modalities, including images, into large language models (LLMs) as a crucial area for progress in artificial intelligence. By introducing multimodal capabilities, these LLMs can enhance the effectiveness of traditional language systems, creating innovative interfaces and experiences while tackling a broader range of tasks. This system card focuses on assessing the safety features of GPT-4V, building upon the foundational safety measures established for GPT-4. Here, we delve more comprehensively into the evaluations, preparations, and strategies aimed at ensuring safety specifically concerning image inputs, thereby reinforcing our commitment to responsible AI development. Such efforts not only safeguard users but also promote the responsible deployment of AI innovations.

ESMC represents the newest advancement in the ESM series of protein language models, pushing the boundaries of representation learning within the field of protein biology. With training on billions of evolutionary sequences, it adeptly captures representations that encapsulate a mechanistic understanding of protein structure and function. The model utilizes a transformer architecture, focusing on sequences as its primary modality, and is trained on a vast dataset comprising up to 6 billion proteins. ESMC is tailored for various protein science applications, such as predicting structures, annotating functions, designing proteins, and exploring evolutionary connections among proteins. Additionally, it possesses the capability to create novel proteins based on partial sequences, structures, or functional constraints, thereby enabling researchers to investigate innovative avenues in protein design and biological discovery. Accessible through the Biohub Platform, ESMC can be utilized via an API and the ESM Python package, which includes quickstart resources for installation, API key generation, and platform connectivity, ensuring a seamless experience for users. This comprehensive accessibility encourages a broader engagement with protein research and enhances collaborative efforts in the scientific community.

MiniMax M2.5 is a next-generation foundation model built to power complex, economically valuable tasks with speed and cost efficiency. Trained using large-scale reinforcement learning across hundreds of thousands of real-world task environments, it excels in coding, tool use, search, and professional office workflows. In programming benchmarks such as SWE-Bench Verified and Multi-SWE-Bench, M2.5 reaches state-of-the-art levels while demonstrating improved multilingual coding performance. The model exhibits architect-level reasoning, planning system structure and feature decomposition before writing code. With throughput speeds of up to 100 tokens per second, it completes complex evaluations significantly faster than earlier versions. Reinforcement learning optimizations enable more precise search rounds and fewer reasoning steps, improving overall efficiency. M2.5 is available in two variants—standard and Lightning—offering identical capabilities with different speed configurations. Pricing is designed to be dramatically lower than competing frontier models, reducing cost barriers for large-scale agent deployment. Integrated into MiniMax Agent, the model supports advanced office skills including Word formatting, Excel financial modeling, and PowerPoint editing. By combining high performance, efficiency, and affordability, MiniMax M2.5 aims to make agent-powered productivity accessible at scale.

Opus, recognized as our most advanced model, surpasses its competitors in numerous widely-used evaluation benchmarks for artificial intelligence, including assessments of undergraduate expert knowledge (MMLU), graduate-level reasoning (GPQA), fundamental mathematics (GSM8K), and others. Its performance approaches human-like comprehension and fluency in handling intricate tasks, positioning it at the forefront of general intelligence advancements. Furthermore, all Claude 3 models demonstrate enhanced abilities in analysis and prediction, sophisticated content creation, programming code generation, and engaging in conversations in various non-English languages such as Spanish, Japanese, and French, showcasing their versatility in communication.

Odyssey-2 represents a cutting-edge interactive video technology that allows for immediate and real-time video generation that users can engage with. Simply enter a prompt, and the system promptly starts streaming several minutes of video that reacts to your input. This innovation transforms video from a traditional playback experience into a responsive, action-sensitive stream: the model operates in a causal and autoregressive manner, crafting each frame based on previous frames and your actions instead of adhering to a set timeline, which enables a seamless adaptation of camera perspectives, environments, characters, and narratives. The platform efficiently begins video streaming nearly instantaneously, generating new frames approximately every 50 milliseconds (around 20 frames per second), ensuring that you don’t have to wait long for content but instead immerse yourself in an evolving narrative. Beneath its surface, the model employs an advanced multi-stage training process that shifts from generating fixed clips to creating open-ended interactive video experiences, granting you the ability to type or voice commands while exploring a world crafted by AI that responds in real-time. This innovative approach not only enhances engagement but also revolutionizes the way viewers interact with visual storytelling.

OpenAI Astra is an upcoming AI model concept built for users and organizations that need more capable multimodal reasoning, autonomous execution, and end-to-end task completion. The model would be designed to understand complex instructions, process multiple types of input, use tools, and produce polished outputs across professional workflows. Astra would support advanced use cases such as software engineering, research automation, document analysis, data interpretation, creative generation, enterprise search, and AI agent orchestration. Its multimodal capabilities would allow users to work across text, images, files, interfaces, and potentially voice-based interactions in a more natural way. For developers, Astra could provide a foundation for building AI-native applications, copilots, assistants, and agent systems. For businesses, it could help automate multi-step work that normally spans teams, tools, and long review cycles. The model would likely emphasize reliability, safety, context retention, and stronger task follow-through. Astra would be especially valuable for workflows where users need not only answers, but completed work that can be inspected, refined, and deployed. By combining frontier reasoning, multimodal intelligence, tool use, and agentic execution, OpenAI Astra would help push AI from conversation toward practical collaboration.

Anthropic has introduced Claude Haiku 4.5, its newest small language model aimed at achieving near-frontier capabilities at a significantly reduced cost. This model mirrors the coding and reasoning abilities of the company's mid-tier Sonnet 4, yet operates at approximately one-third of the expense while delivering over double the processing speed. According to benchmarks highlighted by Anthropic, Haiku 4.5 either matches or surpasses the performance of Sonnet 4 in critical areas such as code generation and intricate "computer use" workflows. The model is specifically optimized for scenarios requiring real-time, low-latency performance, making it ideal for applications like chat assistants, customer support, and pair-programming. Available through the Claude API under the designation “claude-haiku-4-5,” Haiku 4.5 is designed for large-scale implementations where cost-effectiveness, responsiveness, and advanced intelligence are essential. Now accessible on Claude Code and various applications, this model's efficiency allows users to achieve greater productivity within their usage confines while still enjoying top-tier performance. Moreover, its launch marks a significant step forward in providing businesses with affordable yet high-quality AI solutions.

Prime Intellect serves as a comprehensive superintelligence framework, offering a cohesive platform for computation, training, inference, and experimentation for groups aiming to develop, implement, and enhance their models over time. Instead of relying on advancements from frontier models, the stack emphasizes ownership of intelligence, providing users with a singular loop for reinforcement learning environments, extensive training, evaluations, inference, and computing needs. Within the Lab, teams can enable self-improving agents by transforming tasks into reinforcement learning settings and utilizing the Prime CLI for creation, development, evaluation, and deployment. The Environment Hub presents access to an extensive collection of over 2,500 open-source RL environments, while hosted evaluations allow teams to assess model performance across various open-source frameworks without the burden of managing infrastructure. Additionally, Hosted Training facilitates large-scale models tailored for agentic workflows, ensuring managed training processes with complete visibility and control, along with direct assistance from the dedicated applied research team, allowing for a more robust and user-friendly experience in model development. This integrated approach not only streamlines the development process but also fosters innovation and collaboration among teams.

Ling 3.0 Tiny is a reasoning model featuring open weights, comprising 7.9 billion total parameters and 1.3 billion active parameters, alongside a substantial context window of 262,000 tokens. Leveraging a mixture-of-experts architecture, it pushes the boundaries of the open-weights Pareto frontier in terms of intelligence relative to active parameters, while being compact enough for local deployment in various environments. Scoring 25 on the Artificial Analysis Intelligence Index, it stands on par with gpt-oss-120b, which scores 24, despite utilizing 15 times fewer total parameters and 4 times fewer active parameters. This impressive parameter efficiency does come with a trade-off, as it requires a significant 213 million output tokens to complete the Intelligence Index evaluation. In addition, Ling 3.0 Tiny exhibits noteworthy advancements in reducing hallucination tendencies compared to Ling-mini-2.0; it enhances its AA-Omniscience score by 59 points while keeping accuracy levels consistent. Notably, rather than making random guesses in uncertain situations, the model chose to attempt only 37% of the questions during evaluation, leading to a markedly reduced hallucination rate of 30%, a significant improvement over the previous generation's 96%. This strategic approach not only demonstrates the model's improved reasoning capabilities but also highlights its potential for more reliable real-world applications.

Gemini 2.5 Deep Think represents an advanced reasoning capability within the Gemini 2.5 suite, employing innovative reinforcement learning strategies and extended, parallel reasoning to address intricate, multi-faceted challenges in disciplines such as mathematics, programming, scientific inquiry, and strategic decision-making. By generating and assessing various lines of reasoning prior to delivering a response, it yields responses that are not only more detailed and creative but also more accurate, while accommodating longer interactions and integrating tools like code execution and web searches. Its performance has achieved top-tier results on challenging benchmarks, including LiveCodeBench V6 and Humanity’s Last Exam, showcasing significant improvements over earlier iterations in demanding areas. Furthermore, internal assessments reveal enhancements in content safety and tone-objectivity, although there is a noted increase in the model's propensity to reject harmless requests; in light of this, Google is actively conducting frontier safety evaluations and implementing measures to mitigate risks as the model continues to evolve. This ongoing commitment to safety underscores the importance of responsible AI development.

Phi-4 is an advanced small language model (SLM) comprising 14 billion parameters, showcasing exceptional capabilities in intricate reasoning tasks, particularly in mathematics, alongside typical language processing functions. As the newest addition to the Phi family of small language models, Phi-4 illustrates the potential advancements we can achieve while exploring the limits of SLM technology. It is currently accessible on Azure AI Foundry under a Microsoft Research License Agreement (MSRLA) and is set to be released on Hugging Face in the near future. Due to significant improvements in processes such as the employment of high-quality synthetic datasets and the careful curation of organic data, Phi-4 surpasses both comparable and larger models in mathematical reasoning tasks. This model not only emphasizes the ongoing evolution of language models but also highlights the delicate balance between model size and output quality. As we continue to innovate, Phi-4 stands as a testament to our commitment to pushing the boundaries of what's achievable within the realm of small language models.

OpenGPT-X is an initiative based in Germany that is dedicated to creating large AI language models specifically designed to meet the needs of Europe, highlighting attributes such as adaptability, reliability, multilingual support, and open-source accessibility. This initiative unites various partners to encompass the full spectrum of the generative AI value chain, which includes scalable, GPU-powered infrastructure and data for training expansive language models, alongside model design and practical applications through prototypes and proofs of concept. The primary goal of OpenGPT-X is to promote innovative research with a significant emphasis on business applications, thus facilitating the quicker integration of generative AI within the German economic landscape. Additionally, the project places a strong importance on the ethical development of AI, ensuring that the models developed are both reliable and consistent with European values and regulations. Furthermore, OpenGPT-X offers valuable resources such as the LLM Workbook and a comprehensive three-part reference guide filled with examples and resources to aid users in grasping the essential features of large AI language models, ultimately fostering a deeper understanding of this technology. By providing these tools, OpenGPT-X not only supports the technical development of AI but also encourages responsible usage and implementation across various sectors.

Fugu-Ultra v1.1 represents the enhanced multi-agent orchestration model developed by Sakana AI, designed for intricate coding tasks, agentic functions, and sophisticated reasoning capabilities. Instead of depending on a singular model, it adeptly manages a variety of cutting-edge models, strategically selecting and merging specialized agents tailored for individual tasks, all while offering a unified model interface. The latest orchestration upgrade in v1.1 features the inclusion of newer frontier models, resulting in improved performance metrics across all monitored benchmarks, achieving enhancements of up to 7.9 points compared to v1.0, with particularly notable successes on ProgramBench and Terminal Bench 2.1. In addition, Fugu can now seamlessly integrate within Claude Code through endpoints that are compatible, allowing a collaborative ensemble of models to function within the user-friendly terminal environments for coding, debugging, reviewing, and executing scripts. Users can easily set up this integration using a one-command installer on Ubuntu and macOS, while alternative manual configurations are provided for Windows and other systems, ensuring accessibility across diverse platforms. This advancement not only streamlines workflows but also enhances productivity by leveraging the strengths of multiple specialized agents.

Axiomatic AI represents a cutting-edge artificial intelligence platform that aims to enhance the efficiency of scientific research and engineering processes by merging generative AI with rigorous mathematical validation and physics-oriented reasoning. Central to this platform is the principle of Axiomatic Intelligence, which harmonizes advanced AI models with formal logic and specialized world models to guarantee that the generated outputs are both mathematically sound and physically coherent. In contrast to traditional AI systems that may yield plausible yet unverified responses, Axiomatic AI incorporates robust verification mechanisms to ensure that results are measured against established formal specifications and engineering criteria prior to being presented to users. This innovative methodology enables the platform to effectively assist in high-stakes applications across various fields, including but not limited to photonics, electronics, thermal engineering, mechanics, and signal analysis. As a result, Axiomatic AI stands out as a valuable tool for professionals seeking reliable solutions in complex scientific domains.

Nova 2 Pro represents the pinnacle of Amazon’s Nova family, offering unmatched reasoning depth for enterprises that depend on advanced AI to solve demanding operational challenges. It supports multimodal inputs including video, audio, and long-form text, allowing it to synthesize diverse information sources and deliver expert-grade insights. Its performance leadership spans complex instruction following, high-stakes decision tasks, agentic workflows, and software engineering use cases. Benchmark testing shows Nova 2 Pro outperforms or matches the latest Claude, GPT, and Gemini models across numerous intelligence and reasoning categories. Equipped with built-in web search and executable code capability, it produces grounded, verifiable responses ideal for enterprise reliability. Organizations also use Nova 2 Pro as a foundation for training smaller, faster models through distillation, making it adaptable for custom deployments. Its multimodal strengths support use cases like video comprehension, multi-document Q&A, and sophisticated data interpretation. Nova 2 Pro ultimately empowers teams to operate with higher accuracy, faster iteration cycles, and safer automation across critical workflows.

At Liquid, we aim to develop highly advanced AI systems that can address challenges of varying magnitudes, enabling users to construct, utilize, and manage their own AI solutions effectively. This commitment is designed to guarantee that AI is seamlessly, dependably, and efficiently incorporated across all businesses. In the long run, Liquid aspires to produce and implement cutting-edge AI solutions that are accessible to all individuals. Our approach involves creating transparent models within an organization that values openness and clarity. Ultimately, we believe that this transparency fosters trust and innovation in the AI landscape.

Gemini 3 Flash is a next-generation AI model created to deliver powerful intelligence without sacrificing speed. Built on the Gemini 3 foundation, it offers advanced reasoning and multimodal capabilities with significantly lower latency. The model adapts its thinking depth based on task complexity, optimizing both performance and efficiency. Gemini 3 Flash is engineered for agentic workflows, iterative development, and real-time applications. Developers benefit from faster inference and strong coding performance across benchmarks. Enterprises can deploy it at scale through Vertex AI and Gemini Enterprise. Consumers experience faster, smarter assistance across the Gemini app and Search. Gemini 3 Flash makes high-performance AI practical for everyday use.

Upstage has unveiled Solar Pro 2, a cutting-edge large language model designed for frontier-scale applications, capable of managing intricate tasks and workflows in various sectors including finance, healthcare, and law. This model is built on a streamlined architecture with 31 billion parameters, ensuring exceptional multilingual capabilities, particularly in Korean, where it surpasses even larger models on key benchmarks such as Ko-MMLU, Hae-Rae, and Ko-IFEval, while maintaining strong performance in English and Japanese as well. In addition to its advanced language comprehension and generation abilities, Solar Pro 2 incorporates a sophisticated Reasoning Mode that significantly enhances the accuracy of multi-step tasks across a wide array of challenges, from general reasoning assessments (MMLU, MMLU-Pro, HumanEval) to intricate mathematics problems (Math500, AIME) and software engineering tasks (SWE-Bench Agentless), achieving problem-solving efficiency that rivals or even surpasses that of models with double the parameters. Furthermore, its enhanced tool-use capabilities allow the model to effectively engage with external APIs and data, broadening its applicability in real-world scenarios. This innovative design not only demonstrates exceptional versatility but also positions Solar Pro 2 as a formidable player in the evolving landscape of AI technologies.

rready builds sovereign enterprise software for work and innovation management, AI-native, API-first, and hosted entirely in Europe, free from extraterritorial access like the US CLOUD Act. Spun out of Swisscom in 2021 (rready AG, Zurich), it serves 40+ enterprise and public-sector customers including BMW, Tetra Pak, Swisscom, LGT and Implenia. sovara is Europe's sovereign alternative to Jira and Confluence: one platform for software development, project management, issue tracking, and knowledge management. Built for regulated environments, it pairs predictable pricing and granular governance with a modular architecture on sovereign cloud (Swiss/EU) or on-premise, with a guaranteed Jira migration. rready innovate is AI-native innovation management: capture, evaluate, and scale ideas from concept to measurable impact — spanning idea management, intrapreneurship, and continuous improvement. Our mission: give enterprises full ownership of how they manage work and drive innovation, now and in the future.

Ornith-1.0 represents an innovative family of models tailored specifically for coding tasks that require agentic capabilities. This family encompasses a wide range of models, from the compact 9B Dense versions ideal for deployment on edge devices to the expansive 397B MoE frontier-scale models designed for peak performance, including variants such as 9B Dense, 31B Dense, 35B MoE, and 397B MoE. Built upon the foundational strengths of pretrained models like Gemma 4 and Qwen 3.5, Ornith-1.0 excels in achieving top-tier performance among open-source models that are similar in size when evaluated against coding benchmarks. A significant breakthrough of this model is its self-improving training framework, which effectively learns to produce both solution rollouts and the tailored scaffolds that direct those rollouts. Rather than depending on static, human-crafted harnesses, Ornith-1.0 perceives the scaffold as a dynamic entity that evolves alongside the policy, enabling the model to optimize both the orchestration of tasks and the resulting solutions in tandem. This dual optimization approach enhances the model's adaptability and effectiveness in real-world coding scenarios.