Microsoft published documentation for Windows 11 version 26H2 on 30 September. It confirms the annual feature update ships as a small enablement package on the same servicing branch as 25H2 and 24H2, and switches on features previously held behind temporary commercial controls.
AI & models30 September 20267 min readSources 8
On 30 September, Cloudflare open sourced Forge, a generation pipeline that turns an API spec into SDKs, CLIs, docs and MCP servers. NVIDIA also published Kumo Tabular, a foundation model that predicts from tables in one forward pass, with no training or feature engineering. Neither is a fine tuning tool, but both show where the open stack is moving.
AI & models30 September 20269 min readSources 15
Fulcrum released Echo on 30 September, a writing model it says cost less than $5,000 to train and beats frontier models at style imitation. It was one of four small or local-first releases that landed the same day, alongside Phonon-2, Strata and Ollama's decision-model endpoint.
AI & models30 September 20268 min readSources 11
Modal and Carnegie Mellon's Full Stack Data Lab published Quail on 30 September. The inference engine plans SQL queries and schedules LLM execution together, and on a single H100 it clears one billion tokens per minute, more than 10x a vLLM baseline.
AI & models30 September 20263 min readSources 7
Two new inference engines claim order-of-magnitude efficiency gains on the same GPU hardware. Meanwhile, OpenAI's new $500 subscription and a September pricing snapshot show the per-token market still moving in opposite directions.
AI & models30 September 20267 min readSources 10
Modal and Carnegie Mellon's Full Stack Data Lab say their Quail engine pushed a multi-join AI-SQL query past one billion tokens per minute on a single H100. That is more than 10x their vLLM baseline, in a post published on 30 September.
AI & models30 September 20265 min readSources 7