TheZvi diffs OpenAI's new Model Spec: philosophically similar, but new examples dodge hard cases
TheZvi · x · 2026-11-04
OpenAI updated its Model Spec, arguably its most important document defining how its models should think and behave. TheZvi built a Google Doc to diff the versions and concludes:
- Most changes are relocations and line edits, mostly improvements; the new Spec is philosophically and practically similar to the old one
- His main new critique: many of the new examples "do not carve reality at its joints" — they have overdetermined answers and fail to teach the model or reader what to do in genuinely difficult cases
- He focuses on what is actually new and important rather than every edit
A substantive third-party read of how OpenAI's model-behavior governance is evolving.
More from Models
- Gemini 4 Carbon takes #1 on Artificial Analysis Intelligence Index — thursdai_pod · 2026-11-04
- Open decision model REAX 1.3 launches: €0.40 per million input tokens, output free — airesearch12 · 2026-11-04
- REAX 1.3 ships open weights under MIT license with two miner-trained models inside — airesearch12 · 2026-11-04
- Anthropic faces questions over Claude's rules on AI-generated video game clones — emmanuelvivier · 2026-11-04
- Simon Willison replicates GPT-6 Astra weight-interpretation feat with Qwen 3.8 27B in 45 minutes — cephaloform · 2026-11-04
- Qwen, GPT and Gemini Flash all fail the same way on this service — matdryhurst · 2026-11-04