Model Efficiency
High impactLive published free itemOpenAI prices GPT-6 Astra at $10 and $50 per million tokens
GPT-6 Astra lands at $10 per million input tokens and $50 per million output, matching Claude Fable 5.1 base rates. Fast mode costs 2x standard. Subscription users get it inside existing allowances, off by default at launch.
What changed
OpenAI published GPT-6 Astra with API standard pricing of $10 per 1M input and $50 per 1M output tokens, plus an optional Fast mode at up to 2x speed for 2x the standard price. Astra also adds note-preserving context management, so Codex can store and search notes across prior context windows instead of compressing them away.
Why it matters
Frontier per-token pricing at OpenAI and Anthropic is now identical, so price is no longer the lever between them. The lever becomes cost per completed task: how many tokens the model generates, how many tool calls it makes, and how often you rerun the job. Against GPT-5.6 Sol at its current promotional $4 and $20, Astra is 2.5x the input and output rate.
Take
GPT-6 Astra lands at $10 per million input tokens and $50 per million output, matching Claude Fable 5.1 base rates. Fast mode costs 2x standard. Subscription users get it inside existing allowances, off by default at launch.
What to do
Do not move default workloads to Astra on launch. Run one representative task on Astra and on your current model, and compare total cost to a finished, accepted result rather than headline rate. If you are on a subscription, check whether your admin has enabled it before assuming access.
Evidence
No public sources are linked yet.