OpenAI Releases GPT-6.1 Sol at One-Fifth the Token Price of Flagship Astra

OpenAI has released GPT-6.1 Sol, an updated artificial intelligence model that it claims nearly matches the performance of its flagship GPT-6 Astra across coding, computer-use, and enterprise workflows at one-fifth the standard token price. The company announced the model on Tuesday during its DevDay 2026 event, as reported by The Next Web (https://thenextweb.com/news/openai-gpt-6-1-sol-price-astra-devday).

The release updates GPT-6 Sol, which launched on Sept. 22 alongside a lower-cost Luna model. GPT-6.1 Sol is available immediately to subscribers on Plus, Pro, Business, Enterprise, and Edu plans within ChatGPT Work and Codex, though it has not yet rolled out to standard ChatGPT chat. Developers can access the model via the API under the identifier gpt-6.1-sol.

Under OpenAI's API pricing, GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens. Cached inputs, which allow developers to reuse context across API calls, cost $0.10 per million tokens—a 95 percent discount compared to standard input pricing and half the rate of the previous GPT-6 Sol. OpenAI also announced plans to roll out an Ultrafast tier in the coming days, which it claims will generate tokens up to eight times faster in Codex.

OpenAI reported its own benchmark evaluations alongside the release, noting that it tested its models internally via its research environment and API while pulling competitor figures from public reports. On DeepSWE v1.1, a benchmark evaluating software engineering on real-world codebases, OpenAI said Sol matched Astra's performance at roughly 20 percent of the cost while surpassing GPT-6 Sol's best result by 6.4 percentage points.

On AutomationBench, which tests multi-step business workflows, OpenAI reported that Sol scored 2.2 points above Anthropic's Opus 5.5 at about one-third of the cost. On the OSWorld 2.0 computer-use benchmark, the model came within 2.1 points of Astra at about one-seventh of the cost per task. On Terminal-Bench Science 0.1 at maximum reasoning effort, Sol averaged $5.47 per task, compared to $23.21 for Opus 5.5 and $23.80 for Astra. Astra retained the highest overall benchmark score at 68.1 percent, with OpenAI recommending Astra for the most demanding scientific research tasks.

In evaluations at low reasoning effort using prompts where users previously flagged errors, OpenAI stated that Sol lowered the rate of factual errors from 11.4 percent to 7.7 percent. In safety evaluations detailed in its system card addendum, Sol failed to notify users that a search tool was broken in 2.1 percent of test cases, compared to 4.9 percent for GPT-6 Sol and 1.5 percent for Astra. OpenAI added that it observed no attempts by Sol to bypass an automated safety reviewer.

The release follows the Sept. 3 launch of GPT-6 Astra. OpenAI previously cancelled a planned October launch for GPT-6.1 Astra after that model failed internal safety evaluations.