GPT-6.1 Sol: API prices, long-context costs and limits
GPT-6.1 Sol is a paid developer model. OpenAI describes near-Astra performance at lower standard token prices, but the real bill depends on prompt size, output, caching and tools.
Official publisher artwork hosted by its publisher. Official source. Independent guide, not affiliated or sponsored. Artwork and trademarks belong to their owners.
What is GPT-6.1 Sol?
OpenAI presents it for complex coding, computer use and professional work. "Near-Astra" is the provider's performance description, not a guarantee for your project. Compare models on a small set of your real tasks before choosing on price alone.
The model page lists a 1,050,000-token context window and up to 128,000 output tokens. It separately caps input at 922,000 tokens. Tokens are small units of text processing, not the same as words. A large context window means room for input and output under the model's rules; it does not guarantee perfect recall of every detail in a long document.
Standard prices and approximate INR
For prompts up to 272K input tokens, current Standard rates per one million tokens are:
- Input: US$2, approximately ₹193.
- Cached input: US$0.10, approximately ₹9.64.
- Cache writes: US$2.50, approximately ₹241.
- Output: US$10, approximately ₹964.
These illustrations use the reported 6 October exchange rate of approximately ₹96.43 per dollar. They exclude taxes, card conversion costs and separate service charges. Caching is not free: cache reads and writes have their own rates.
If a prompt exceeds 272K input tokens, the higher rates apply to the full request: input and cache rates double, while output costs 1.5 times the normal rate. Long-context rates per million tokens are US$4 input (about ₹386), US$0.20 cached input (about ₹19.29), US$5 cache writes (about ₹482) and US$15 output (about ₹1,446). It is not only the tokens beyond the threshold that move to the higher rate.
The model page says Fast mode costs twice Standard; Batch and Flex are 50% lower than Standard. Ultrafast mode costs six times Standard. Regional processing adds 10% where offered. Tool calls and supporting services can add charges. Do not use a token-only estimate as an all-in quote.
Supported features and limits
The page lists text input and output plus image input, but no audio or video modality support. It says to use the Responses API for tool calling; Chat Completions is supported without tool calling. Listed Responses tools include file search, code interpreter, hosted shell, computer use and MCP.
Reasoning effort supports low, medium, high, xhigh and max; medium is the default. None and minimal are not supported. The current API rate table lists paid usage tiers; no free API allowance was established. A ChatGPT subscription should not be treated as prepaid API usage or proof that this exact model is available in that plan.
How to try it with controlled spending
- Read the model and pricing pages. Choose the API route and tools needed for your task.
- Set provider-side spending alerts or limits and your own request caps. Keep the API key out of browser frontend code and public repositories.
- Start with short, non-sensitive examples and limited output. Inspect usage and the cost breakdown.
- Test accuracy, latency and failure cases against another suitable model. Evaluate Malayalam names and numbers separately if relevant.
- Trim unnecessary history before a prompt crosses 272K. Check cached-token accounting rather than assuming all repeated text is discounted.
- Give tools only the permissions needed. Review generated code, external messages and file changes before using them.
Safety and use in India
OpenAI's system card treats GPT-6.1 Sol as Critical capability in cybersecurity under its own Preparedness Framework, with safeguards. That is a capability classification, not permission to attack systems or a promise that all requests will be accepted. Use security tools only within authorised scope.
Indian developers should review data-handling requirements before sending client code or personal data. The page describes US and EU residency support, not an India processing region. Do not assume API processing happens in India just because you pay in rupees.
What changed since the original post?
The 1 October post gave the main input/output prices. This guide adds cache-write costs, the full-request long-context multiplier, endpoint-specific tool rules and processing premiums. INR examples are updated; API billing remains separate from ChatGPT plan claims.
Frequently asked questions
Is it free?
No free API allowance was established in the checked current model and pricing pages.
Does only the excess above 272K cost more?
No. The model page says the higher rates apply to the full request.
Was a benchmark run here?
No. This guide verifies published model terms and prices, not independent performance.
Sources: Official model page · Official pricing · System card · Current ChatGPT help · INR conversion reference. Checked 7 October 2026.
Scoped recheck 24 October 2026: official model, API pricing, system card and current ChatGPT help checked. Standard and full-request long-context prices remain supported. Added 922,000 maximum input tokens and Ultrafast 6x Standard premium. Current API table lists paid usage tiers; no free allowance established. ChatGPT Work/Codex rollout does not include prepaid API billing. Older labelled INR examples unchanged. Official artwork retained; no billed request, account entitlement, benchmark or India checkout test.