OpenAI announced GPT-6.1 Sol on Sept. 29, 2026, listing API rates of $2 per million standard input tokens, $0.10 per million cached input tokens and $10 per million output tokens. Its standard input and output rates are one-fifth of GPT-6 Astra’s listed rates; cached input costs one-tenth of Astra’s listed cached-input rate. These are API token prices, not monthly ChatGPT subscription charges.
OpenAI launches GPT-6.1 Sol for coding and computer use
OpenAI described GPT-6.1 Sol as an upgrade to GPT-6 Sol for agentic coding, computer use and complex professional work. Developers can access it through the API using the model identifier gpt-6.1-sol.
OpenAI Developers’ launch post described the model’s intended work, including complex code refactors and long-running agents. The post also highlighted the reduced cached-input rate.
GPT-6.1 Sol API pricing by token category
OpenAI listed these API rates for GPT-6.1 Sol and GPT-6 Astra:
| Model | Standard input per 1 million tokens | Cached input per 1 million tokens | Output per 1 million tokens |
| GPT-6.1 Sol | $2 | $0.10 | $10 |
| GPT-6 Astra | $10 | $1 | $50 |
The “one-fifth” price comparison applies to standard input and output: GPT-6.1 Sol’s listed rates in those categories are one-fifth of GPT-6 Astra’s. Cached input is a separate category, priced at one-tenth of Astra’s listed rate. OpenAI also said GPT-6.1 Sol’s cached-input rate is 50% lower than GPT-6 Sol’s listed cached-input rate of $0.20 per million tokens.
What OpenAI’s benchmark results show
OpenAI reported that GPT-6.1 Sol matched GPT-6 Astra on DeepSWE v1.1 at roughly one-fifth the cost per task. That result applies to this coding benchmark, not to every task or workload.
On the OSWorld 2.0 offline set at maximum reasoning effort, OpenAI reported that GPT-6.1 Sol scored 2.1 percentage points below Astra at roughly one-seventh the cost per task. On Terminal-Bench Science 0.1, also at maximum effort, OpenAI reported an average task cost of $5.47 for GPT-6.1 Sol and $23.80 for Astra; Astra’s reported score of 68.1% was the highest among the models tested.
OpenAI also reported that responses containing factual errors fell from 11.4% for GPT-6 Sol to 7.7% for GPT-6.1 Sol at low reasoning effort. That evaluation used deliberately difficult prompts drawn from user-flagged factual-error conversations, which OpenAI said were not representative of typical use.
Launch access and the planned Codex Ultrafast option
At launch, OpenAI listed GPT-6.1 Sol access through the API and through ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. OpenAI said the model was not yet available in Chat at launch.
OpenAI also announced a planned Codex Ultrafast option, describing token generation of up to eight times standard speed. The launch announcement said the option would arrive “in the coming days.”
A nine-test run adds a separate datapoint
A nine-test Stonelabs evaluation reported a weighted score of 96 for GPT-6.1 Sol. The reported cost for the complete suite was $2.63 for Sol, about $14 for GPT-6 Astra and about $12 for Opus 5.5. The run also included a failure on the penguin task. These are whole-suite test costs, not API rates per million tokens.
