Anthropic Claude 4.1 Opus: First Model to Pass Internal 10-Hour Autonomous Software Engineering Test
Claude 4.1 Opus became the first publicly known frontier model to successfully complete a full 10-hour autonomous software engineering task (building a functional web app from natural language spec with zero human intervention) in Anthropic’s internal red-team evaluation suite.
new
Mistral NeMo 2.0: 12B Model with State-of-the-Art On-Device Multimodal Performance
Anthropic Claude 4.1 Opus: First Model to Pass Internal 10-Hour Autonomous Software Engineering Test
Google Gemini Robotics 1.0: End-to-End Multimodal Policy Model for Dexterous Manipulation
DeepSeek-V3.5: 671B MoE Model Surpasses GPT-5.2 on Chinese & English Long-Context Benchmarks
OpenAI o4-mini Pro Launches with Breakthrough Chain-of-Verification Reasoning
Mistral Large 2.1: Open Model with 128K Context and Strong Reasoning in Non-English Languages
xAI Grok-3.5 Early Access: Enhanced Long-Context Reasoning and Native Tool Integration
Meta Llama 4 Scout Released: 405B Parameters with Strong Multimodal and Tool-Use Capabilities
Google DeepMind AlphaFold 3.1 Update: Improved Protein-Ligand Binding Prediction Accuracy
Anthropic Claude 4 Sonnet: New Multi-Agent Collaboration Framework for Complex Tasks