GPT-6.1 Sol

GPT-6.1 Sol, released September 29, 2026, was the first point release of OpenAI's GPT-6 generation, arriving one week after GPT-6 Sol and Luna and nearly four weeks after GPT-6 Astra opened the line. OpenAI's documentation positioned it against the flagship rather than against the Sol it replaced: "near-Astra performance at a lower cost for complex coding, computer use, and professional work", with a suggestion that developers run it beside Astra on their own tasks to judge the tradeoff between quality and cost. The list price did not move from GPT-6 Sol's: $2.00 per million input tokens and $10.00 per million output, a fifth of Astra's $10.00 and $50.00. Cached input fell to $0.10 per million tokens, half of GPT-6 Sol's cached rate.

OpenAI's launch post charted every score against cost per task. On DeepSWE 1.1 it scored 75.2% at release, 6.4 points above GPT-6 Sol's best and level with Astra at roughly a fifth of the cost. On the OSWorld 2.0 offline set it reached 71.4% at maximum effort, seven points above GPT-6 Sol and 2.1 below Astra at about a seventh of the cost per task. It scored 36.2% on AutomationBench, 32.0% on GDP.pdf and 57.1% on Terminal-Bench Science 0.1, more than double GPT-6 Sol's maximum-effort score there at $5.47 per task, against $23.80 for Astra, which had the top score in OpenAI's comparison at 68.1%. OpenAI also reported fewer factual errors on difficult prompts, 7.7% against GPT-6 Sol's 11.4% at low effort, and no attempts to bypass its automated safety reviewer.

The API surface kept the family's 1,050,000-token context window, 128,000 max output tokens, text and image input and text output, with a knowledge cutoff of April 30, 2026, ten days later than GPT-6 Sol's. The reasoning-effort control narrowed at the bottom: low, medium, high, xhigh and max were supported with medium as the default, while the none setting GPT-6 Sol had offered was dropped and minimal was not supported. Tool calling ran through the Responses API, with Chat Completions supported only without tools, and the model shipped with both US and EU data residency, Fast mode being unavailable under the EU option. It was served in the API as gpt-6.1-sol.