GPT-6.2 Sol
A general-purpose model in OpenAI's GPT-6.2 series aimed at everyday coding, writing, and agent workflows, supporting the construction of interactive web applications from a brief.
Core Facts
The API model identifier for GPT-6 Astra is gpt-6-astra.
GPT-6.2 Astra and Sol cover the three major scenarios of ChatGPT Work, Codex and API at lower cost.
OpenAI's GPT-6.2 series includes two separate models: Astra and Sol, both launched on AI Gateway.
Timeline (last 90 days)
In OpenAI's LiveWork test, GPT-6.2 Sol outperformed the older GPT-6 Astra, at a task cost of about one-seventh of the latter's.
According to confirmation by user Teebo, GPT-6.2 Sol's Ultra Fast mode has been opened to the $200-per-month ChatGPT Pro plan; OpenAI says Token speed in this mode can reach up to 4.5 times that of the standard mode, with plan quota consumption calculated at 6 times the standard tier.
In the Yorkshire transforming-into-mecha anime test, GPT-6.2 Astra and Sol both underperform Claude Fable 5.2 and Gemini 4 Carbon at the "Yorkshire transforms into a mecha" step.
The Databricks platform announced that Astra and Sol, two new models in the GPT-6.2 series, are now officially available.
YouTube's official demo video, in the form of a workplace scenario drama, shows the model handling multiple tasks such as meetings, financial updates and presentation creation, generating workbooks in real time via voice commands, updating accounts payable and tracking metrics.
Currently the public information is mostly demo-based, lacking technical details, context window specifications, pricing and availability information.
In the demo the model output the conclusion: "We'd owe 32 million more in 2031, but next year, we make 30 cents less per share".
The instruction in the screenshot, "Track nap debt through 2031 and show next year's EPS, both against plans", shows that the model was asked to handle cross-year financial forecasting and comparative analysis.
The instructions in the demo are fragmented and colloquial, interspersed with interruptions and repetitions, and the model needs to maintain task continuity amid this chaos.
Both GPT-6.2 Astra and Sol support fast mode and OpenAI's Ultrafast service tier, which trades a higher per-token price for lower latency.
12 more timeline events
All Facts (20)
The API model identifier for GPT-6 Astra is gpt-6-astra.
90%Partially VerifiedGPT-6.2 Astra and Sol cover the three major scenarios of ChatGPT Work, Codex and API at lower cost.
75%Partially VerifiedOpenAI's GPT-6.2 series includes two separate models: Astra and Sol, both launched on AI Gateway.
65%UnverifiedAccording to confirmation by user Teebo, GPT-6.2 Sol's Ultra Fast mode has been opened to the $200-per-month ChatGPT Pro plan; OpenAI says Token speed in this mode can reach up to 4.5 times that of the standard mode, with plan quota consumption calculated at 6 times the standard tier.
50%UnverifiedIn OpenAI's LiveWork test, GPT-6.2 Sol outperformed the older GPT-6 Astra, at a task cost of about one-seventh of the latter's.
50%UnverifiedIn the Yorkshire transforming-into-mecha anime test, GPT-6.2 Astra and Sol both underperform Claude Fable 5.2 and Gemini 4 Carbon at the "Yorkshire transforms into a mecha" step.
50%UnverifiedIn the demo the model output the conclusion: "We'd owe 32 million more in 2031, but next year, we make 30 cents less per share".
50%UnverifiedThe instruction in the screenshot, "Track nap debt through 2031 and show next year's EPS, both against plans", shows that the model was asked to handle cross-year financial forecasting and comparative analysis.
50%UnverifiedCurrently the public information is mostly demo-based, lacking technical details, context window specifications, pricing and availability information.
50%UnverifiedThe instructions in the demo are fragmented and colloquial, interspersed with interruptions and repetitions, and the model needs to maintain task continuity amid this chaos.
50%UnverifiedSelection rule: whether the task spans multiple steps, whether it involves modifying an existing codebase, and whether it requires strictly following a template or operating a computer interface; lean toward "yes" and choose Astra, lean toward "no" and choose Sol.
50%UnverifiedGPT-6.2 Sol targets everyday coding, writing, and agent workflows, and the official announcement specifically mentions that it can build an interactive web app from a written brief.
50%UnverifiedBoth GPT-6.2 Astra and Sol support fast mode and OpenAI's Ultrafast service tier, which trades a higher per-token price for lower latency.
50%UnverifiedIn this release announcement, OpenAI did not elaborate on the specific differences between the Astra and Sol versions, but instead presented the two as a single generation of models.
50%UnverifiedOpenAI's official announcement is brief and does not disclose the GPT-6.2 model's parameter scale, context window, specific benchmark results or pricing details.
50%UnverifiedOpenAI officially stated that the new GPT-6.2 model has improved efficiency and lower cost.
50%UnverifiedGemini 4 Hydrogen's cost per single task is about 32% of Sonnet 5.5's, but is still about 2.5 times more expensive than GPT-6.1 Sol.
50%UnverifiedThe Databricks platform announced that Astra and Sol, two new models in the GPT-6.2 series, are now officially available.
50%UnverifiedYouTube's official demo video, in the form of a workplace scenario drama, shows the model handling multiple tasks such as meetings, financial updates and presentation creation, generating workbooks in real time via voice commands, updating accounts payable and tracking metrics.
50%UnverifiedOpenAI has released a usage guide for the GPT-6 series, introducing how to choose among the three models—Astra, Solo, and Luna—based on task requirements, and how to adjust reasoning levels and speed modes.
50%