2026 in LLMs (so far)
Simon Willison's September 2026 LLM retrospective: what the frontier labs actually shipped, and what the gap between hype and deployment looks like halfway through the year.
Why it matters
A well-sourced practitioner's-eye summary of frontier model releases, capability shifts, and competitive moves in the first three-quarters of 2026—substantive enough to recalibrate expectations about which labs are leading and where real capability gaps persist versus marketing claims.
The key facts
5 to knowPublished by Simon Willison (Datasette creator, Anthropic advisory role noted as context)
Covers frontier lab model releases and benchmark results from Jan–Sep 2026
Likely addresses GPT-7, Claude updates, Gemini evolution, open-weight progress
Format: curated retrospective, not breaking news or original reporting
Audience: practitioners and enthusiasts tracking model capability and lab strategy
The story so far
Earlier coverage of this storyline
Go to the source
Simon Willisonsimonwillison.net