Overview
GPT-6.2 Sol is the everyday model OpenAI shipped on November 18, 2026, alongside the GPT-6.2 Astra flagship. The name comes second, but this is the tier most requests will actually hit. The pricing decision is straightforward: token rates are identical to GPT-6.1 Sol at $2 input and $10 output, with more capability on top. Existing users can move up without touching their budget.
The benchmark position is more interesting than the price. On FrontierCode 2.0 it sits within 0.5 points of the Astra flagship at roughly a fifth of the cost. LifeVal comes in above GPT-6 Astra, the previous flagship, at about a seventh of the cost. GDPval v2 improves 5.3 points over GPT-6.1 Sol, and AA-3D puts it third at 1,259 Elo. Read together, those numbers say the gap between Sol and Astra has shrunk to the point where paying four times more for it rarely makes sense on ordinary work.
Sol also gets a launch perk Astra does not. Ultrafast reaches Pro 200 subscribers for the first time, starting with this model. Ultrafast is 50% cheaper for three months and costs API and Enterprise users 3x standard. For a daily driver, response speed shapes the experience more than peak capability does, so this handoff matters more in practice than a few benchmark points.
One thing to note: it carries the same Critical cybersecurity rating as Astra. That is OpenAI's own system card grade, applied to both models and printed in the release notes, which makes it a neutral data point rather than a warning. The launch page also omits the context window and maximum output token count, and no third-party parameter table fills them in, so those two fields should stay empty rather than guessed.
Choosing between the two is simple. Daily work, code, research cleanup, routine agent pipelines, use Sol. Switch to Astra only for genuinely hard projects that span a dozen tools and run for half an hour or more. The signal is output length and tool-call count per run, not how impressive the task sounds.
Key Features
- Same price, more capability: $2 input, $10 output and $0.10 cached per million tokens, identical to GPT-6.1 Sol. GDPval v2 improves 5.3 points, so existing users upgrade without a budget change.
- Within 0.5 points of the flagship: FrontierCode 2.0 lands within 0.5 points of GPT-6.2 Astra at roughly a fifth of the cost. Paying four times more for half a point rarely pays off on everyday coding.
- LifeVal above the last flagship: LifeVal scores higher than GPT-6 Astra at about a seventh of the cost. Last generation's flagship capability now sits at this generation's mid tier.
- Ultrafast reaches Pro 200 first here: Pro 200 subscribers get Ultrafast for the first time, beginning with Sol. It runs 50% cheaper for three months and costs API and Enterprise users 3x standard.
- Third on AA-3D: 1,259 Elo on AA-3D for third place, in the same band as the flagship's 1,318. 3D and design work does not automatically need the top tier.
- All four subscription tiers on day one: Live in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu, with the API model name gpt-6.2-sol. OpenRouter lists it at the same $2 and $10.
Use Cases
- Everyday driver: code, research cleanup and routine agent pipelines where this tier covers most requests
- Cost-sensitive high volume: $2 input with $0.10 cache reads for backend jobs running tens of thousands of times a day
- Frontend and product prototypes: FrontierCode 2.0 is nearly level with the flagship while interfaces need fast responses
- Pro 200 subscribers who want speed: first access to Ultrafast at that tier
- 3D and design assistance: third on AA-3D, good enough at a fifth of the flagship's price
- Smooth upgrades from GPT-6.1 Sol: identical token rates, no budget rework and no code changes
Pros
- Token rates match GPT-6.1 Sol exactly, so the capability gain costs nothing extra
- FrontierCode 2.0 within 0.5 points of the flagship at a fifth of the cost is a well-placed everyday value
- LifeVal beats the previous flagship GPT-6 Astra at about a seventh of the cost
- Ultrafast reaches Pro 200 and starts with Sol, lowering the speed floor for subscription tiers
- Third on AA-3D and in the flagship's band, so 3D work does not require an automatic upgrade
- ChatGPT Work, Codex and the API all live on launch day, and existing users change one model name
Pricing
API standard rates: $2 per million input tokens, $10 per million output, $0.10 for cached input, matching GPT-6.1 Sol. Ultrafast: 50% off for three months after launch, charged at 3x standard for API and Enterprise users. Pro 200 subscribers get it for the first time starting with this model; Pro 500 can use it on both. Subscriptions: ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu, with Enterprise and Edu requiring an admin to enable it. Channels: OpenAI API under the model name gpt-6.2-sol, ChatGPT and Codex; OpenRouter at the same rates. For reference: the Astra flagship is $8 and $40 with $0.40 cached.
Summary
GPT-6.2 Sol is OpenAI's mid-tier everyday model from November 18, 2026, priced level with GPT-6.1 Sol at $2 input, $10 output and $0.10 cached. It lands within 0.5 points of the Astra flagship on FrontierCode 2.0 at about a fifth of the cost, beats the previous flagship GPT-6 Astra on LifeVal, improves 5.3 points over GPT-6.1 Sol on GDPval v2, and ranks third on AA-3D at 1,259 Elo. Ultrafast reaches Pro 200 subscribers for the first time starting with this model, at 50% off for three months. Cybersecurity is rated Critical, the same band as Astra, and no context window or maximum output token count has been published.
Version History
- GPT-6.2 Sol launch (2026-11-18): Ships alongside the Astra flagship with token rates level with GPT-6.1 Sol: $2 input, $10 output, $0.10 cached. Within 0.5 points of Astra on FrontierCode 2.0 at a fifth of the cost, and above GPT-6 Astra on LifeVal. Ultrafast reaches Pro 200 for the first time, starting here.