Microsoft Introduces Open-Source Tool for AI Behavior Testing
Microsoft has unveiled Adaptive Spec-driven Scoring for Evaluation and Regression Testing (ASSERT), an open-source framework enabling developers to cr...
TechCrunch AI
datasette-agent-micropython 0.1a0 Released for Safe Python Code Execution
Simon Willison announced the alpha release of datasette-agent-micropython 0.1a0, a tool aiming to enable safe generation and execution of Python code ...
Simon Willison
JetBrains Unveils Mellum2: A New 12B Mixture-of-Experts Model
JetBrains has introduced Mellum2, a 12B Mixture-of-Experts model, marking a new development in the field of large language models. Further specifics o...
Hugging Face
Mistral AI Launches ‚Mistral 3‘ Model
Mistral AI has announced the launch of Mistral 3, signifying a new generation or major update to their flagship AI model. This release is expected to ...
Mistral AI
Mistral AI Releases New ‚Mistral Medium 3.5‘ Model
Mistral AI has announced the release of Mistral Medium 3.5, indicating an update to their suite of AI models. This new iteration likely brings perform...
Mistral AI
Mistral AI Introduces ‚Mistral Small 4‘
Mistral AI has unveiled Mistral Small 4, introducing another new model to its growing portfolio. This release suggests a focus on providing efficient ...
Mistral AI
A Beginner’s Guide to Profiling in PyTorch with torch.profiler
Details in article.
Hugging Face
Claude Opus 4.8 Touted for Improved ‚Honesty‘ in Responses
Anthropic is emphasizing the enhanced 'honesty' of its new Claude Opus 4.8 model, stating that it's trained to avoid making unsupported claims. This a...
The Verge AI
Anthropic Unveils Claude Opus 4.7 with Enhanced Performance
Anthropic has released Claude Opus 4.7, its latest foundational AI model, promising stronger performance across coding, agent capabilities, vision, an...
Anthropic
Shipping Trillion-Parameter Models with Delta Weight Sync in TRL
Details in article.
Hugging Face
New ITBench-AA Benchmark Reveals Frontier Models Struggle with Agentic Enterprise IT Tasks
Artificial Analysis and IBM have introduced ITBench-AA, the first benchmark specifically designed for agentic enterprise IT tasks. Initial results sho...
Hugging Face
Human Archive Taps India’s Gig Economy for Robotics Training Data
Human Archive, a startup founded by UC Berkeley and Stanford researchers, is leveraging India's gig economy to collect crucial physical training data ...
TechCrunch AI