Average Ratings 0 Ratings
Average Ratings 1 Rating
Description
Beam represents Reflection’s inaugural open-weight model, characterized as a sparse Mixture-of-Experts framework featuring a staggering 501 billion parameters, of which 23 billion are actively utilized, specifically crafted for tasks involving coding, reasoning, and agentic functions. Its prowess is derived from extensive pretraining and reinforcement learning, having been developed using a vast dataset of 23.8 trillion diverse, high-quality tokens sourced from the web, public domains, and proprietary licensed materials. With a targeted emphasis on enhancing coding and agentic capabilities, Beam is engineered to provide competitive open-weight performance while ensuring efficient inference compute. This model adeptly handles a wide array of tasks, including intricate software engineering, terminal operations, STEM activities, web searches, tool utilization, and general knowledge inquiries. Reinforcement learning techniques have been employed to bolster its abilities in multi-step reasoning, tool application, and responsiveness to environmental feedback. Moreover, users have the flexibility to manage the balance between efficiency and performance through an adjustable reasoning effort parameter, thus tailoring the model's output to better suit their specific needs.
Description
SWE-2 Flash is a high-speed AI software engineering model developed by Cognition for automated coding, development workflows, and technical problem-solving. The model is based on Inkling-Small from Thinking Machines and uses a mixture-of-experts architecture containing 276 billion total parameters and 12 billion active parameters. Cognition applies large-scale mid-training and reinforcement learning to improve the model's programming accuracy, reasoning capabilities, and ability to complete complex software engineering tasks. SWE-2 Flash supports code generation, repository maintenance, automated code reviews, feature implementation, documentation updates, and agent-driven software development. Its optimized architecture delivers approximately 200 tokens per second, making it three times faster than SWE-2 while operating at one-fifth the inference cost. The model achieved a score of 24.0% on FrontierCode 2.0, 45.5% on FrontierCode 1.1 Main, and 89.2% on Terminal-Bench 2.1 in Cognition's published evaluations. Training improvements also target general reasoning, problem-solving, and cybersecurity tasks to expand the range of activities supported by autonomous coding agents. SWE-2 Flash is available in Devin Desktop and CLI, allowing developers to incorporate the model into interactive and automated software engineering workflows. Cognition is also developing a controlled model-weight access program for trusted industry partners and researchers interested in fine-tuning, infrastructure control, and deployment in sensitive environments.
API Access
Has API
No
API Access
Has API
No
Integrations
.NET
No
C#
No
CSS
No
Cerebras
No
Dart
No
Devin Desktop
No
Go
No
HTML
No
JavaScript
No
Kotlin
No
Integrations
.NET
Yes
C#
Yes
CSS
Yes
Cerebras
Yes
Dart
Yes
Devin Desktop
Yes
Go
Yes
HTML
Yes
JavaScript
Yes
Kotlin
Yes
Pricing Details
No price information available.
Free Trial
No
Free Version
No
Pricing Details
$20/month
Free Trial
No
Free Version
Yes
Deployment
Web-Based
Yes
On-Premises
Yes
iPhone App
No
iPad App
No
Android App
No
Windows
Yes
Mac
Yes
Linux
Yes
Chromebook
No
Deployment
Web-Based
Yes
On-Premises
No
iPhone App
No
iPad App
No
Android App
No
Windows
No
Mac
No
Linux
No
Chromebook
No
Customer Support
Business Hours
No
Live Rep (24/7)
No
Online Support
Yes
Customer Support
Business Hours
No
Live Rep (24/7)
No
Online Support
Yes
Types of Training
Training Docs
Yes
Webinars
No
Live Training (Online)
Yes
In Person
No
Types of Training
Training Docs
Yes
Webinars
No
Live Training (Online)
No
In Person
No
Vendor Details
Company Name
Reflection
Country
United States
Website
reflection.ai/blog/introducing-beam
Vendor Details
Company Name
Cognition
Founded
2023
Country
United States
Website
cognition.com