GPT-6 Astra API pricing is $10.00/1M input and $50.00/1M output per 1M tokens on the standard tier. Cached input is $1.00/1M. Cache writes are $12.50/1M. Rates in TokenCalculator checked 17 Sep 2026.
Price your own prompt
Sticker rates are only the start. Paste a real prompt into the cost calculator with GPT-6 Astra selected, then set output size, cache hit rate, and monthly volume.
GPT-6 Astra rate card
API model id: gpt-6-astra. Context window: about 1.05M tokens. Max output: 128K. Exact OpenAI encodings apply in the TokenCalculator tokenizer.
| Line | USD per 1M tokens |
|---|---|
| Input (<=272K prompt) | $10.00/1M |
| Cached input (<=272K) | $1.00/1M |
| Cache write | $12.50/1M |
| Output (<=272K) | $50.00/1M |
| Batch input | $5.00/1M |
| Batch output | $25.00/1M |
| Long-context input (>272K, whole request) | $20.00/1M |
| Long-context output (>272K, whole request) | $75.00/1M |
| Long-context cached input (>272K) | $2.00/1M |
The 272K long-context cliff
When prompt tokens cross 272K, OpenAI reprices the entire request, not only the overflow. Input and cache rates double. Output rises to 1.5x. Design agents around that threshold before you treat the full 1.05M window as a flat price.
Fast mode and Batch
Batch and Flex are about half of standard rates when latency can wait. Fast mode (formerly Priority) is about 2x standard for lower latency. Confirm the service tier on OpenAI’s pricing page before you ship a production path.
How Astra compares to nearby tiers
| Model | Input / 1M | Output / 1M | Cache read / 1M |
|---|---|---|---|
| GPT-6 Astra | $10.00/1M | $50.00/1M | $1.00/1M |
| GPT-5.6 Sol | $4.00/1M | $20.00/1M | $0.400/1M |
| Claude Fable 5.1 | $10.00/1M | $50.00/1M | $0.250/1M |
| Claude Opus 5 | $5.00/1M | $25.00/1M | $0.500/1M |
Astra lists at the same $10 / $50 sticker as Claude Fable 5.1. The real bill gap is cache reads ($1 vs $0.25) and Astra’s long-context surcharge. Versus GPT-5.6 Sol, Astra is a premium step for the hardest end-to-end work.
Worked examples
- Short chat: 2K fresh input + 500 output about $0.045 on standard rates.
- Cached agent turn: 20K cached + 2K fresh + 1K output about $0.071 (cache read $1/M).
- Long prompt over 272K: use the long-context band for the whole request, not only the tokens past the threshold.
When to use Astra
Use Astra when GPT-5.6 Sol still fails your evals on hard coding or agent jobs and you can pay list rates plus Fast mode when needed. Prefer Sol, Terra, or Luna for volume. Prefer Fable 5.1 when stable prefixes dominate and Anthropic cache math wins.
Common mistakes
- Shopping only on the $10 / $50 sticker and ignoring cache and long-context lines
- Filling most of the 1M window without checking the 272K cliff
- Comparing Astra input list price to Opus 5 without equalizing output and cache mix
- Treating curated TokenCalculator rates as a live OpenAI price websocket
Frequently asked questions
How much does GPT-6 Astra cost?
Standard API pricing is $10.00/1M per 1M input tokens and $50.00/1M per 1M output tokens. Cached input is $1.00/1M. Confirm on OpenAI’s pricing page for your account tier.
What is the GPT-6 Astra API model id?
Use gpt-6-astra in the OpenAI API.
Does Astra have a long-context surcharge?
Yes. Prompts over 272K input tokens reprice the full request at higher input, cache, and output rates.
Is Fast mode more expensive?
Yes. Fast mode is about 2x standard rates for lower latency. Batch and Flex are about half when delayed processing is acceptable.
How does Astra compare to Claude Fable 5.1 on price?
Same sticker $10 / $50. Fable 5.1 cache reads are $0.25 vs Astra $1. Astra adds a long-context cliff above 272K. Read the Astra vs Fable guide for workload math.
Is GPT-5.6 Sol cheaper than Astra?
Yes on list rates. Sol is $4.00/1M / $20.00/1M in this catalog. Use Sol when quality is enough.
Are these rates live?
No. TokenCalculator curates official list prices and stamps lastVerified (17 Sep 2026). Confirm critical budgets on the provider page.
Where should I start in TokenCalculator?
Open the cost calculator with GPT-6 Astra preselected, paste a representative prompt, set output tokens and monthly volume, then compare Sol and Fable 5.1 on the same text.
Next steps
- Astra vs Claude Fable 5.1
- Claude Fable 5.1 pricing
- OpenAI API pricing hub
- How we verify rates
- Official OpenAI pricing
Related Sep 2026 pricing guides
Frontier rate cards, cost-per-task math, and long-context cliffs. Cross-link these when you compare Sol, Luna, Opus 5.5, Grok 4.7, or Astra.