Estimate GPT-6.1 Terra, GPT-6.1 Sol, GPT-6 Astra, GPT-6 Sol, GPT-6 Luna and GPT-5.6 API costs, including long-context pricing, 30-minute and 12-hour cache writes, cache reads and output tokens. Enter representative per-request token counts and choose Standard, Batch/Flex, Fast or supported Ultrafast processing. The calculator applies OpenAI’s higher long-context matrix automatically above 272,000 input tokens.
Use token counts from a representative API response wherever possible. Model and processing-mode availability may differ between accounts and regions. Pricing can change; confirm material budgets against OpenAI’s official API pricing and GPT-6 Astra model page, plus the GPT-6 Sol and GPT-6 Luna model pages.
Estimate your OpenAI API cost
Existing model rates and processing modes checked: 19 October 2026. GPT-6.1 Terra added and checked: 27 October 2026. Published API rates in USD.
12-hour cache-write support checked: 23 October 2026. This estimates Responses API text-token charges, not Decisions API billing.
GPT-6 Astra · Standard · Short context · 1,000 requests
- Uncached input
- $10.00
- 30-minute cache writes
- $0.00
- 12-hour cache writes
- $0.00
- Cache reads
- $0.00
- Output
- $25.00
- Average per request
- $0.035
Planning estimate only. It excludes tools, storage, fine-tuning, regional charges, retries, taxes, contracted discounts and non-token fees.
Current OpenAI API rates used by this calculator
Rates below are USD per million tokens for Standard API requests. Batch and Flex processing cost 50% of Standard; Fast processing costs twice the applicable Standard rate. Ultrafast costs six times Standard and is available here only for GPT-6 Astra and GPT-6.1 Sol. A request with more than 272,000 total input tokens uses the long-context rates for the full request.
| Model and context | Input | Cache read | 30-minute cache write | Output |
|---|---|---|---|---|
| GPT-6 Astra – short | $10 | $1 | $12.50 | $50 |
| GPT-6 Astra – long | $20 | $2 | $25 | $75 |
| GPT-6.1 Sol: short | $2 | $0.10 | $2.50 | $10 |
| GPT-6.1 Sol: long | $4 | $0.20 | $5 | $15 |
| GPT-6.1 Terra: short | $1 | $0.05 | $1.25 | $5 |
| GPT-6.1 Terra: long | $2 | $0.10 | $2.50 | $7.50 |
| GPT-6 Sol: short | $2 | $0.20 | $2.50 | $10 |
| GPT-6 Sol: long | $4 | $0.40 | $5 | $15 |
| GPT-6 Luna: short | $0.10 | $0.01 | $0.125 | $0.50 |
| GPT-6 Luna: long | $0.20 | $0.02 | $0.25 | $0.75 |
| GPT-5.6 Sol – short | $4 | $0.40 | $5 | $20 |
| GPT-5.6 Sol – long | $8 | $0.80 | $10 | $30 |
| GPT-5.6 Terra – short | $2 | $0.20 | $2.50 | $12 |
| GPT-5.6 Terra – long | $4 | $0.40 | $5 | $18 |
| GPT-5.6 Luna – short | $0.20 | $0.02 | $0.25 | $1.20 |
| GPT-5.6 Luna – long | $0.40 | $0.04 | $0.50 | $1.80 |
“Long” means more than 272,000 input tokens in one request, not 272,000 tokens across all monthly requests. The table shows default 30-minute writes at 1.25 times the corresponding uncached-input rate. Optional 12-hour Responses API writes cost 2.5 times that input rate, twice the table’s write column. Cache-read rates do not change with lifetime. GPT-5.6 Sol’s listed rates are OpenAI’s promotional rates, available through at least 21 November 2026.
How to estimate an OpenAI API bill
- Choose the exact model. Check the full generation and model name. GPT-6 Sol and Luna have different prices from GPT-5.6 Sol and Luna.
- Choose the processing mode. Standard is the baseline. Batch/Flex halves the applicable rate, Fast doubles it and supported Ultrafast multiplies it by six.
- Enter per-request token counts. Keep uncached input, 30-minute writes, 12-hour writes, cache reads and output separate. For Responses API usage, the two write-lifetime counts sum to cache-write tokens. Subtract reads and both write counts from total input to obtain uncached input; do not count them twice.
- Enter request volume. For a monthly forecast, use the number of calls expected in a typical month.
- Check the context notice. The calculator totals all input categories per request and switches the entire request to the published long-context matrix above 272,000 tokens.
- Reconcile with actual usage. Re-run the estimate when prompts, caching or models change.
Formula used
Total = requests × processing multiplier × ((uncached input × input rate) + (30-minute writes × input rate × 1.25) + (12-hour writes × input rate × 2.5) + (cache reads × cached rate) + (output × output rate)) / 1,000,000
Which OpenAI model should you price?
- GPT-6.1 Terra has short-context Standard rates of $1 input and $5 output per million tokens. Use your own task results to compare it with other models.
- GPT-6 Astra is the higher-priced option for demanding reasoning and agent workflows.
- GPT-6.1 Sol is a separate model with $2 input, $0.10 cache reads, $2.50 default 30-minute cache writes and $10 output per million tokens at short-context Standard rates. Do not substitute GPT-6 Sol cache-read pricing.
- GPT-6 Sol supports coding and agent workflows at $2 per million short-context input tokens and $10 per million output tokens.
- GPT-6 Luna supports focused, high-volume tasks at $0.10 per million short-context input tokens and $0.50 per million output tokens.
- GPT-5.6 Sol is an advanced option for difficult workloads and currently has promotional API pricing.
- GPT-5.6 Terra is the balanced option for production agents and general-purpose work.
- GPT-5.6 Luna is the cost-sensitive option for high-volume or simpler workloads.
Model choice should be based on your own quality, latency and reliability tests. A cheaper rate does not guarantee a lower end-to-end cost if a workflow needs more retries or tokens.
Frequently asked questions
Are GPT-6 Astra, Sol and Luna available through the API?
OpenAI lists gpt-6-astra, gpt-6-sol and gpt-6-luna as API models. Sol and Luna were released on 22 September 2026. Check model access and processing-mode availability in your own account before budgeting a deployment.
Are ChatGPT subscriptions included?
No. ChatGPT plans and API usage are billed separately. This tool estimates API token charges only.
Why are cache writes separate from cached input?
Creating a reusable cache and reading from that cache have different rates. Combining them can materially understate a workload that frequently creates new cache entries.
What is the difference between Standard, Batch/Flex, Fast and Ultrafast?
Standard uses the published base rates. Batch and Flex processing are priced at 50% of Standard, while Fast processing is twice the applicable Standard rate. Ultrafast is six times Standard for GPT-6 Astra and GPT-6.1 Sol, with separate rate limits. Availability and latency characteristics differ, so choose the mode your workload actually uses. GPT-6 Astra Fast is unavailable with EU data residency. GPT-6 Sol and GPT-6 Luna support EU data residency with Standard, Fast, Flex and Batch processing, subject to endpoint and account eligibility. See the official Fast-mode compatibility guidance. GPT-6.1 Sol supports US and EU data residency, including Fast and Ultrafast. Astra Ultrafast supports US and global processing, not EU residency. The 10% regional-processing surcharge is excluded from this calculator.
Does this cover the Decisions API?
No. This calculator estimates Responses API text-token charges. The Decisions API uses different billing, including no output-token charge. Check the official Decisions API pricing and scope for that endpoint.
Why might the final bill differ?
Retries, tool calls, hidden system or tool-schema tokens, model changes, storage, regional pricing, contract terms and taxes can all change the final invoice.
Sources and scope
GPT-6.1 Terra was released on 26 October 2026. Its token rates, cache-write lifetimes, 272,000-input-token boundary and Standard, Batch/Flex and Fast processing were checked on 27 October against the official model page, pricing tables and release changelog. Ultrafast is not offered for Terra here. Account access, quality and latency were not tested.
OpenAI added 12-hour Responses API prompt caching on 20 October 2026. The separate 30-minute and 12-hour write rates and mixed-lifetime accounting were checked on 23 October against the official prompt caching guide and current pricing tables. Use billed token categories from a representative response; cache reuse and latency are not guaranteed.
GPT-6.1 Sol pricing, its 272,000-input-token boundary and supported processing modes were checked on 19 October 2026 against the official GPT-6.1 Sol model page, API pricing tables and Ultrafast availability guide. The older GPT-6 Sol remains a separate option. This is a token-cost estimate, not a live availability or latency test.
Compare providers with the Claude API cost calculator and Gemini API cost calculator, or visit the AI cost calculators hub.