GPT-6.1 Sol closes on Astra at a fifth of the price

OpenAI released GPT-6.1 Sol, an upgrade to GPT-6 Sol that it positions as near Astra intelligence on agentic coding, computer use and professional work, at roughly one-fifth of Astra’s token price. The framing throughout the announcement is cost per task rather than a new capability ceiling, and Astra still holds the top score on the hardest benchmark.

The API price is $2 per million input tokens, $10 per million output, and $0.10 per million cached input. That cached rate is the change most likely to move real bills: OpenAI says it is 95% below standard input pricing and 50% below GPT-6 Sol’s cached rate, which matters for agents that resend the same context across many requests. The model ID is gpt-6.1-sol.

What the benchmarks say

• DeepSWE v1.1: matches Astra at about one-fifth the cost, and beats GPT-6 Sol’s best score by 6.4 points at a lower reasoning effort.
• GDP.pdf: higher than Opus 5.5 with fallbacks at less than half the cost per task.
• AutomationBench 1.0.6: 2.2 points above Opus 5.5 at medium effort at about a third of the cost, and up 4.8 points on GPT-6 Sol at the same setting.
• OSWorld 2.0 offline set: 7 points above GPT-6 Sol at maximum effort, landing within 2.1 points of Astra at roughly one-seventh the cost per task.
• Terminal-Bench Science 0.1: more than doubles GPT-6 Sol at maximum effort. $5.47 per task on average, against $23.21 for Opus 5.5 and $23.80 for Astra. Astra still scores the highest at 68.1%.

The factuality result is the one worth separating from the rest, because OpenAI states the conditions plainly. On conversations where a user had already flagged an earlier model’s factual error, the share of answers containing at least one error drops from 11.4% to 7.7% at low reasoning effort, a roughly 32% reduction. OpenAI says explicitly that these are deliberately hard prompts and are not representative of typical use.

On safety, GPT-6.1 Sol fails to disclose a broken search tool in 2.1% of cases, against 4.9% for GPT-6 Sol and 1.5% for Astra, with no attempts to bypass an automated reviewer in either.

Availability is Plus, Pro, Business, Enterprise and Edu in ChatGPT Work and Codex, plus the API. It is not in Chat. An Ultrafast variant with up to 8x faster token generation in Codex is promised for the coming days.

(Original source: OpenAI)