The Pulse
OpenAI Prices GPT-6.1 Sol at One-Fifth of GPT-6 Astra
OpenAI says GPT-6.1 Sol approaches GPT-6 Astra on several agentic tasks at one-fifth of the standard input and output token prices. The company reports gains in coding, computer use and professional work, while saying Astra remains its reco
AI.info Team · Written with AI ·
GPT-6.1 Sol approaches GPT-6 Astra on several agentic tasks, OpenAI says, while costing one-fifth as much per standard input and output token. But the company’s own results show a trade-off: Astra still posts the top score in its scientific-workflow test, and OpenAI recommends it for the hardest research tasks.
Announced September 29, GPT-6.1 Sol is an upgrade to GPT-6 Sol aimed at coding, computer use and professional work. OpenAI’s comparisons suggest it could bring Astra-like performance to some lower-cost workflows, though the company’s benchmarks—not independent evaluations—are the evidence behind those claims.
Sol closes gaps on coding and computer use
On DeepSWE v1.1, which tests software engineering in real codebases, OpenAI says the new model matches Astra at roughly one-fifth of the cost. It also beats GPT-6 Sol’s best score by 6.4 percentage points while using less reasoning effort and money, according to the announcement.
On OSWorld 2.0’s offline computer-use tasks, GPT-6.1 Sol scores seven percentage points above GPT-6 Sol at maximum reasoning effort for less than half the cost. OpenAI says it comes within 2.1 points of Astra on the same test at roughly one-seventh of Astra’s cost per task.
The results are closer on some professional-work measures. At medium reasoning effort on AutomationBench, Sol scores 2.2 points above Anthropic’s Opus 5.5 at about a third of the cost, and 4.8 points above GPT-6 Sol. On GDP.pdf, a test using complex professional documents, OpenAI says it beats Opus 5.5 with fallbacks at less than half the cost per task and approaches Astra at roughly one-fifth the cost.
Astra keeps the lead on the toughest science test
OpenAI reports an average cost of $5.47 per task for GPT-6.1 Sol on Terminal-Bench Science 0.1 at maximum reasoning effort, compared with $23.21 for Opus 5.5 and $23.80 for Astra. Astra nevertheless records the highest score in that evaluation, 68.1%; OpenAI says it remains the choice for the most difficult scientific research tasks.
The benchmark distinction matters: lower task costs and near-matching scores do not mean the models are interchangeable for every job. OpenAI’s Astra announcement prices that model at $10 per million input tokens and $50 per million output tokens, the basis for GPT-6.1 Sol’s one-fifth comparison.
The lower price comes with limited availability
GPT-6.1 Sol’s standard API rates are $2 per million input tokens, $10 per million output tokens and $0.10 per million cached input tokens. OpenAI says the cached-input rate is 95% below standard input pricing and half the cached-input price for GPT-6 Sol.
The model is available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, as well as through the API under the identifier gpt-6.1-sol. It is not yet available in regular ChatGPT chat. OpenAI says a Codex Ultrafast option, promising up to eight times the standard token-generation speed, will follow in the coming days.
OpenAI flags limits in its own evaluations
OpenAI says GPT-6.1 Sol reduces factual errors on a deliberately difficult set of prompts from 11.4% with GPT-6 Sol to 7.7% at low reasoning effort, a drop of about 32%. The company cautions that the prompts came from de-identified ChatGPT conversations where users had flagged earlier factual mistakes, so the set does not represent typical usage.
The company also reports improved alignment results over GPT-6 Sol, including better performance on tests of user intent and safety constraints. OpenAI says it observed no attempts to bypass an automated safety reviewer, matching results for both GPT-6 Sol and GPT-6 Astra. The release notes that tests were run in OpenAI’s research environment or API, where outputs may differ from production ChatGPT.