OpenAI announced GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026. A coding test published the next day by AI Coding Daily reported that GPT-6.2 Sol medium averaged 5.5 minutes per prompt, compared with 2.5 minutes for GPT-6.1 Sol medium, across the creator’s 35 prompts in eight projects. The result puts a concrete time measurement alongside the launch claims, but describes that test’s workload and setting. OpenAI’s announcement positions Astra for complex professional projects and Sol for everyday work.

When comparing coding models, keep task time and benchmark scores in view alongside token prices: each describes a different part of the work.

What OpenAI announced about Astra and Sol

OpenAI announced GPT-6.2 Astra and GPT-6.2 Sol for ChatGPT Work, Codex, and its API. The company said ChatGPT Work and Codex access was available to Plus, Pro, Enterprise, and Edu users; Enterprise and Edu workspaces require an administrator to enable access. OpenAI describes Astra as its model for complex professional projects and Sol as an efficient, general-purpose option for everyday work.

For API use, OpenAI lists standard prices per million tokens of $8 for Astra input, $0.40 for cached input, and $40 for output. Sol costs $2 per million input tokens, $0.10 for cached input, and $10 per million output tokens. The total for a coding task depends on how many input and output tokens it uses. For background on Sol’s previous generation, see our coverage of GPT-6.1 Sol.

What OpenAI’s Astra demonstration shows

Watch OpenAI’s room-reconstruction and Blender demonstration

OpenAI’s demonstration shows GPT-6.2 Astra in Codex reconstructing a room from two video frames, building the scene in Blender, then adjusting its decor and lighting. The workflow ends with an interactive space and a virtual webcam background.

What AI Coding Daily’s test measured

Watch AI Coding Daily’s 35-prompt coding test

AI Coding Daily reported an average of 5.5 minutes per prompt for GPT-6.2 Sol medium, versus 2.5 minutes for GPT-6.1 Sol medium, in its 35-prompt test across eight projects. The creator also reported that GPT-6.2 Astra scored roughly one point below GPT-6 Astra at high effort on the channel’s leaderboard. AI Coding Daily said benchmark runs were slow and that not every effort-level run was complete when the video was recorded.

OpenAI’s own benchmark claims concern different evaluations: the company said Astra scored 5.2 points above GPT-6 Astra on Agents’ Last Exam at less than one-third the cost. OpenAI also reported a 5.5-point lead over Claude Fable 5.2 at 5% of its cost on that evaluation; its note excludes nine fallback-model tasks from Fable 5.2’s leaderboard result. These are OpenAI’s reported results, while the timing figures above belong to AI Coding Daily’s coding test.

The test’s reported medium-effort timing was 5.5 minutes per prompt for GPT-6.2 Sol and 2.5 minutes for GPT-6.1 Sol; its separate high-effort leaderboard result put GPT-6.2 Astra roughly one point below GPT-6 Astra.