GPT-6 Astra
Language Available Full comparison ↗ openai/gpt-6-astra · by OpenAI
· decoder-only-transformer
Pricing — 1 offering(s)
Input tokens
- $10.00 / 1M tokens (input)
Current
2026-09-24 → present
SCO-639: Standard processing, short-context rate. Also on OpenAI's pricing page but NOT tracked as tiers…
SCO-639: Standard processing, short-context rate. Also on OpenAI's pricing page but NOT tracked as tiers here: cached input $1.00/1M and cache writes $12.50/1M (the schema has no cached-input unit, and a separate per_1m_tokens_input tier would be read as the headline input price by routing and the site; see the cached-input support card), long- context (>272K input tokens) $20.00/1M input and $75.00/1M output, and Batch/Flex at 50% of Standard. No promotional-pricing caveat shown.
Output tokens
- $50.00 / 1M tokens (output)
Current
2026-09-24 → present
SCO-639: Standard processing, short-context rate. See the input tier's note for the cached-input…
SCO-639: Standard processing, short-context rate. See the input tier's note for the cached-input, long-context and Batch/Flex rates that aren't tracked here.
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
See how GPT-6 Astra fits into a cost-aware routing setup
See how →Capability profile
How GPT-6 Astra rates across core capability dimensions, with the task-level evidence behind each rating.
1,050,000-token context window, 128,000 max output tokens (developers.openai.com/api/docs/models/gpt-6-astra).
OpenAI API docs: supports Functions, Web search, File search, Computer use.
Highest-priced tier in the GPT-6 family ($10/$50 per 1M tokens, Standard, short context).
Ratings are estimated — limited independent data is available for this model.
Operator guidance
Top of the GPT-6 family. Route here only when the task needs OpenAI's most capable model; GPT-6 Sol is one fifth the price for balanced intelligence/cost, and GPT-6 Luna covers cost-sensitive, high-volume work.
Use cases
- Hardest end-to-end work: complex reasoning, coding, computer use, research and document creation, per OpenAI's model docs
Limitations
- SCO-639: reasoning, coding, instruction-following, multilingual and speed ratings are left out. No benchmark or per-tier claim was sourced for this entry, and none is inferred.
- Capability ratings are qualitative, not a single cited benchmark run
Citations
- OpenAI API pricing
- OpenAI models: GPT-6 Astra
- GPT-6 Astra: A new generation of intelligence — Announcement, 2026-09-03 (title and date via OpenAI RSS; the page itself returns 403 to automated fetches)