OpenAI released GPT-6.1 Sol on September 29, 2026, one week after GPT-6 Sol. It costs $2 per million input tokens and $10 per million output tokens, one-fifth of GPT-6 Astra's standard rates, and OpenAI says it comes close to Astra on agentic coding, computer use and professional work. The cached-input price of $0.10 per million tokens is the change most likely to move agent budgets.
OpenAI's GPT-6 lineup has three tiers: GPT-6 Astra at the top, GPT-6.1 Sol in the middle, and GPT-6 Luna for high-volume, low-cost work. OpenAI's model guide for the GPT-6 family, published October 2, 2026, frames the choice as balancing capability, cost and latency through three settings: the model, the reasoning effort and the speed tier.
GPT-6.1 Sol is the newer successor to GPT-6 Sol, which OpenAI introduced on September 22, 2026. The announcement keeps the same $2 input and $10 output prices and halves cached input from GPT-6 Sol's rate.
| Per million tokens (USD) | GPT-6.1 Sol | GPT-6 Astra | GPT-6 Luna |
|---|---|---|---|
| Input | $2.00 | $10.00 | $0.10 |
| Cached input | $0.10 | $1.00 | $0.01 |
| Cache writes | $2.50 | $12.50 | $0.125 |
| Output | $10.00 | $50.00 | $0.50 |
Sources: OpenAI model pages for GPT-6.1 Sol, GPT-6 Astra and GPT-6 Luna, checked October 10, 2026.
The GPT-6.1 Sol model page lists the multipliers that sit on top of those rates:
Cache behavior matters as much as cache price. OpenAI's prompt caching update from September 22, 2026 says the GPT-6 family gives cache discounts for eligible shared prefixes reused within a 30-minute window, and adds a dashboard and a diagnostics field that reports why a request missed the cache (for example, changed tool definitions). For an agent that resends the same instructions and tools on every turn, keeping that prefix stable is what turns the $0.10 rate into a real saving.
On October 8, 2026, OpenAI's release notes made Ultrafast mode generally available for GPT-6.1 Sol in the Responses API. It is requested with service_tier: "ultrafast" and reduces the time between generated output tokens. It is the same model on a faster serving tier, priced at 6x standard, with US and EU data residency available. That price makes it a tool for latency-critical paths, such as a user waiting on a streamed answer, rather than for batch work.
OpenAI's announcement reports results on DeepSWE v1.1, GDP.pdf, AutomationBench, OSWorld 2.0 and Terminal-Bench Science 0.1. The pattern across them is consistent: GPT-6.1 Sol scores well above GPT-6 Sol and approaches Astra at a fraction of the cost per task, while Astra keeps the top score on the hardest work. On Terminal-Bench Science, OpenAI reports an average cost of $5.47 per task for GPT-6.1 Sol at maximum effort, against $23.80 for Astra and $23.21 for Claude Opus 5.5, and says Astra should still be used for the most difficult scientific tasks.
These figures are OpenAI's. The announcement states that competitor results came from public reports and that its own evaluations ran in its research environment or API. This article does not reproduce them. As with Claude Opus 5.5, the vendor is competing on cost per completed task, which is the measure worth testing on your own work.
On safety, OpenAI says GPT-6.1 Sol improves on GPT-6 Sol in its alignment evaluations, and publishes the details in a GPT-6.1 Sol system card addendum.
Teams on GPT-6 Sol get the same input and output prices with cheaper cached input. Re-test prompts and effort settings, since OpenAI reports different behavior at each effort level.
Teams running routine agent work on Astra have the clearest case to test a move down a tier. Measure success rate and cost per completed task, and keep Astra for the tasks where Sol falls short.
Teams comparing vendors can set GPT-6.1 Sol against Claude Sonnet 5.5, which also lists at $2 input and $10 output per million tokens per Anthropic's Sonnet 5.5 announcement and pricing documentation, with a $0.10 cache-read rate since October 7, 2026, a cut announced in the Claude Haiku 5.5 post. See the Claude Sonnet 5.5 guide. The headline prices match; the differences are in caching rules, long-context pricing and per-task token use.
gpt-6.1-sol.service_tier: "ultrafast" only where output latency matters.When was GPT-6.1 Sol released?
September 29, 2026. Ultrafast mode became generally available on October 8, 2026.
How much does GPT-6.1 Sol cost?
$2.00 input, $0.10 cached input, $2.50 cache writes and $10.00 output per million tokens at standard rates, per OpenAI's model page (checked October 10, 2026). Requests above 272,000 input tokens are billed at higher rates for the whole request.
What is the context window of GPT-6.1 Sol?
1,050,000 tokens, with up to 128,000 output tokens and an April 30, 2026 knowledge cutoff, per the model page.
Is GPT-6.1 Sol the same as GPT-6 Sol?
No. GPT-6.1 Sol is an upgrade released a week after GPT-6 Sol, with the same input and output prices and a lower cached-input price.
Discover more content: