Top story 1Anthropic has published a public GitHub repository containing its implementation of Agent Skills for Claude, with instructions, scripts, and resources that the model loads dynamically to perform specialized tasks repeatably across creative, technical, and enterprise workflows. The repository also points to the broader Agent Skills standard at agentskills.io, signaling a push toward reusable, portable agent capabilities rather than one-off prompts. This could accelerate standardization and make enterprise agent deployments easier to audit, share, and reproduce.
G GitHubTop story 2Ant Ling has launched Ling-3.2-flash, a native multimodal mixture-of-experts model with 560B total parameters, 25B activated parameters, and a 1M-token context window supporting text, image, and video inputs. Reported benchmark scores include MathVision 96.80, OmniDocBench 1.5 92.21, and AndroidWorld 85.63, alongside demonstrations of PDF-to-code, GUI operation, 3D reconstruction, and multi-agent workflows. Both OpenAI- and Anthropic-compatible API access are documented, positioning the release for long-context multimodal agent applications.
r RedditTop story 3Nace released Drex Max v1.1, a multimodal decision model it says achieves state-of-the-art results and sits on the Pareto frontier according to Artificial Analysis. The company also published training details in a paper, made previous open weights available on Hugging Face, and plans v1.1 weights with local Mac inference via Kapitsa; the model supports a roughly 256K context window. The claim adds competitive pressure in the emerging multimodal decision-model segment, where benchmark standing and deployability both matter.
r Reddit