OpenAI has upgraded both its flagship and mid-tier reasoning models today, releasing GPT-6.2 Astra and Sol for ChatGPT Work and Codex. The update trims Astra’s API input and output prices by 20% compared with GPT-6 Astra, introduces an Ultrafast generation mode for Pro subscribers, and delivers coding improvements that early testers say result in cleaner code with fewer revision cycles, according to 9to5Mac.
Key takeaways
- GPT-6.2 Astra and GPT-6.2 Sol are live today in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu subscribers.
- Both models complete demanding tasks with fewer output tokens, stretching usage allowances further.
- Astra’s API pricing falls 20% versus GPT-6 Astra; Sol retains its GPT-6.1 pricing.
- A new Ultrafast mode generates tokens at up to 4.5× the standard rate; the $200/month Pro plan gets it on Sol, the $500/month plan on both models.
- On Cognition’s FrontierCode 2.0 benchmark, Sol reached within half a point of Astra at approximately one-fifth of Astra’s cost.
- API access for developers is available immediately alongside the consumer rollout.
- From GPT-6 to GPT-6.2: A Quick Timeline
- What GPT-6.2 Astra Adds for Creative and Coding Work
- GPT-6.2 Sol: Near-Flagship Performance at a Fraction of the Price
- GPT-6.2 Pricing and Plan Availability
- Ultrafast Mode: Up to 4.5× Standard Generation Speed
- What This Means for Developers and Enterprise Teams
- Frequently Asked Questions
- The Bottom Line
From GPT-6 to GPT-6.2: A Quick Timeline
OpenAI launched the GPT-6 generation with Astra in September 2026, then brought Sol up to GPT-6.1 at DevDay a few weeks later. Today’s GPT-6.2 release advances both models simultaneously — an unusually paired upgrade that signals OpenAI is treating Astra and Sol as a coordinated product line rather than two separate roadmaps. The core pitch for both models is efficiency: OpenAI says they finish demanding tasks with fewer output tokens, which means users can accomplish more before reaching their plan’s usage ceiling.
What GPT-6.2 Astra Adds for Creative and Coding Work
Astra remains OpenAI’s recommended model for the most demanding tasks. On the creative side, OpenAI says GPT-6.2 Astra can take a screenshot of a single slide and use it as a design template — matching layout, colours, and font sizes across an entire presentation. That kind of visual-to-structured-output capability extends the model beyond text generation into something closer to a design assistant operating from a visual reference.
For software development, OpenAI reports that GPT-6.2 Astra is better at making and testing changes inside an existing codebase. Early testers cited cleaner code style and more focused tests, with fewer rounds of revision needed to reach a mergeable state. For teams using Codex on long-running projects, that reduction in back-and-forth is a practical time saving. Developers working with AI coding agents will find both improvements directly relevant to their daily workflows.
GPT-6.2 Sol: Near-Flagship Performance at a Fraction of the Price
Sol is positioned as the more affordable option for everyday coding and writing, and GPT-6.2 pushes its performance closer to Astra than the price gap might suggest. OpenAI points to Cognition’s FrontierCode 2.0 benchmark, which measures whether code changes to real-world repositories are good enough for maintainers to accept. With both models running at maximum reasoning effort, Sol came within half a point of Astra — while costing roughly one-fifth as much.
That cost differential matters significantly in API-heavy workflows. Using our AI API cost calculator, teams can model exactly how much that gap translates to at their own token volumes. For context, our AI models database lists GPT-6 Sol at $2.00 per million input tokens and $10.00 per million output tokens — Sol’s pricing carries forward to GPT-6.2, while Astra’s reduction brings it down from the GPT-6 Astra rate of $10.00 in / $50.00 out to a 20% lower figure.
GPT-6.2 Pricing and Plan Availability
| Model | API Input (per 1M tokens) | API Output (per 1M tokens) | Change vs previous |
|---|---|---|---|
| GPT-6 Astra (previous) | $10.00 | $50.00 | — |
| GPT-6.2 Astra | ~$8.00 | ~$40.00 | −20% |
| GPT-6.1 Sol (previous) | $2.00 | $10.00 | — |
| GPT-6.2 Sol | $2.00 | $10.00 | Unchanged |
The 20% reduction on Astra’s API pricing applies to standard input and output rates. Sol’s pricing is unchanged from GPT-6.1. Both models are available today through the API, and both are live in ChatGPT Work and Codex for subscribers on Plus, Pro ($200/month and $500/month tiers), Enterprise, and Edu plans. For teams evaluating overall cost-effectiveness against competing models, the AI price-performance index provides a broader comparison framework.
Ultrafast Mode: Up to 4.5× Standard Generation Speed
The headline feature for Pro subscribers is Ultrafast mode, which OpenAI says generates tokens at up to 4.5 times the standard rate on either model. The trade-off is that it consumes more of a plan’s usage allowance per session. Access is tiered: the $200/month Pro plan receives Ultrafast on Sol only, while the $500/month Pro plan gets Ultrafast on both Astra and Sol. Enterprise and Edu workspaces gain access once an administrator enables it. Users can toggle between speeds from the model picker inside ChatGPT.
The practical value of Ultrafast mode depends heavily on the task. For iterative coding sessions where a developer is waiting on output between edits, a 4.5× speed increase can meaningfully compress the feedback loop. For longer analytical tasks where the bottleneck is thinking time rather than token output, the speed gain matters less and the allowance cost is harder to justify.
What This Means for Developers and Enterprise Teams
The simultaneous release to both the consumer ChatGPT Work interface and the developer API is deliberate: OpenAI is signalling that GPT-6.2 is production-ready from day one rather than a staged rollout. For enterprise teams already using Codex for automated code review or generation pipelines, the efficiency improvements and lower Astra pricing reduce the cost of running high-volume tasks without requiring a change in model selection.
The FrontierCode 2.0 result for Sol deserves particular attention for teams watching their API spend. A model that performs within half a point of a flagship at approximately one-fifth of the cost — without sacrificing context length (both models carry 1.05M token context windows, per our database) — is a meaningful signal for anyone currently defaulting to Astra out of habit rather than necessity. The AI price-performance index offers a structured way to run that comparison against the broader market.
Frequently Asked Questions
What is GPT-6.2 Astra? GPT-6.2 Astra is OpenAI’s updated flagship reasoning model, available today in ChatGPT Work and Codex. It improves on GPT-6 Astra with better visual templating, cleaner code generation in existing projects, and a 20% reduction in API pricing.
How does GPT-6.2 Sol compare to Astra on coding benchmarks? According to OpenAI, GPT-6.2 Sol scored within half a point of GPT-6.2 Astra on Cognition’s FrontierCode 2.0 benchmark at maximum reasoning effort, at roughly one-fifth of Astra’s cost.
Who gets Ultrafast mode and what does it do? Ultrafast mode generates tokens at up to 4.5 times the standard rate. The $200/month Pro plan gets it on Sol; the $500/month Pro plan gets it on both Astra and Sol. Enterprise and Edu admins can enable it for their workspaces. It draws from your usage allowance faster than standard mode.
Has GPT-6.2 Sol’s API pricing changed from GPT-6.1 Sol? No. Sol retains its GPT-6.1 pricing. Only Astra’s API rates have been reduced, by 20% compared with GPT-6 Astra.
Is GPT-6.2 available on the API today? Yes. OpenAI confirmed that both GPT-6.2 Astra and GPT-6.2 Sol are accessible via the API as of November 17, 2026, alongside the ChatGPT Work and Codex rollout.
The Bottom Line
GPT-6.2 Astra and Sol represent a tightly focused efficiency upgrade rather than a wholesale architectural change. The 20% price cut on Astra’s API makes the flagship model more accessible for high-volume production use, while Sol’s near-parity performance on FrontierCode 2.0 at a fraction of the cost gives cost-conscious teams a compelling reason to reconsider their default model choice. Ultrafast mode adds a genuine speed lever for Pro subscribers, though it comes at the expense of allowance headroom. For developers and enterprise teams already embedded in the OpenAI ecosystem, both models are worth evaluating immediately — the rollout is live, the API is open, and the pricing changes apply from today.
Sources: 9to5mac.com. Reported November 17, 2026.
