GPT 6 Astra lets you change reasoning effort mid-conversation without breaking the cache
intellectronica · x · 2026-09-05
Per this tweet, GPT 6 Astra has a notably practical capability: you can change reasoning effort mid-conversation without invalidating the context cache.
- The workflow: start on Astra light (roughly equivalent to Sol medium), then raise thinking effort on the fly if the task turns out to benefit from more;
- Key value: no need to rewrite the full context cache when switching, balancing cost and flexibility for long conversations.
Note: this comes from an individual post and is an unconfirmed report on model behavior.
More from Models
- Gemini somehow can't read its own Personal Instructions, user's toggles don't help — Connect_Reporter_799 · 2026-10-23
- Diffusion LLM bias rankings shift with sampling configs, persist after debiasing — Bollegala · 2026-10-23
- Running ~100 DeepSeek v4.1 flash agents in parallel for under $20/month — zack_overflow · 2026-10-23
- Researcher laments open-source audio models are surprisingly subpar, even representation ones — gowthami_s · 2026-10-23
- User claims they generated video with "Claude Opus 5.5", unannounced model name surfaces — jwt0625 · 2026-10-23
- Security researchers push back as OpenAI restricts Astra access; BlackHat 2027 mocked as 'cyber model' bazaar — cyb3rops · 2026-10-23