Banger paper from Stanford and Together AI. They show why it might be a good idea to let your agent team learn its own way of working together. (bookmark it) Three models (o3-mini, Claude Sonnet 4 and DeepSeek-V3) averaged 66.7% across five math and physics benchmarks as a self-organizing team. Their strongest member alone scored 48.8%, and a perfect router choosing among the members' independent answers scored 59.0%. On AIME 2026 the team reached 71.2%, 13.4 points above that router. One member reviews the team's earlier exchanges and rewrites the teamwork strategy, covering roles, the order of discussion phases, who participates and how partial answers are combined. The strategies were learned from only 15 AIME 2024 problems and then applied unchanged to held-out AIME 2024 problems and four new benchmarks. Paper:

GLM-5.3 has arrived on Together AI on Day 0 @Zai_org pushed the GLM-5.2 base further with scaled post-training, delivering stronger coding and long-horizon agent performance. Across benchmarks, it nearly matches Fable 5 while costing a fraction as much per task. Try now:

SF Tech Week meets Fleet Week ✈️ Oct 9: Together AI, @dtcp_capital, and @nvidia are taking over a private waterfront chalet for a front row seat to the Blue Angels. Custom aviators. Great food and drinks. Nonstop action. Speed, precision, peak performance. Fitting theme for the crew in this room. RSVP:

This Week in AI: 🔵 Unified Qualcomm #Dragonwing# stack brings together AI tools, runtimes, and hardware to simplify development and deployment of production-ready Edge AI applications: 🔵 In a live AMA, @durga_malladi, EVP and GM, Technology Planning, Edge Solutions & Data Center at Qualcomm, joined @Modular co-founders Chris Lattner and Tim Davis to discuss the acquisition and the evolving developer roadmap: 🔵 @AIatMeta's open-weight Muse Glimmer LLM is optimized for @Snapdragon platforms, helping developers build personalized, always-on agentic AI experiences on device:

Qwen3.8-2.4T-A95B is live on Together AI. The Qwen team’s new open-weight MoE packs 2.4T total parameters with 95B active, a 1M context window, and strong coding & agent capabilities. Start building: