GPT-6.2 Sol Explained: What It Is, Pricing and How It Compares to Astra

Short answer: GPT-6.2 Sol is OpenAI's mid-priced GPT-6.2 model, released on November 17, 2026 alongside the pricier GPT-6.2 Astra. Reported API pricing is $2 per million input tokens and $10 per million output tokens (the same as GPT-6.1 Sol), with a context window of about 1.05 million tokens. Astra scores higher on tests but costs about four times as much per token.

What GPT-6.2 Sol is

Sol is the workhorse tier of OpenAI's lineup, and the 6.2 version succeeds GPT-6.1 Sol. According to a detailed write-up from eesel AI that cites OpenAI's model page and launch post, the API model ID is gpt-6.2-sol, the knowledge cutoff is June 1, 2026, and it is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex as well as through the API. Enterprise and Edu workspaces need an admin to switch it on. Sol also has a few quirks: the API rejects a reasoning effort of "none", and the lowest setting is "low".

Pricing

Per million tokens, for prompts up to 272,000 input tokens, as reported by AlphaCorp and eesel AI:

  • Standard: $2 input, $0.10 cached input, $10 output.
  • Batch: $1 input and $5 output (eesel AI).
  • Prompts above 272,000 input tokens cost more for the whole request: $4 input and $15 output.
  • Faster modes cost extra: Fast is $4 / $20, and Ultrafast is $6 / $30 through February 28, 2027 before rising.

VentureBeat's comparison table, written for Google's rival launch, lists the same $2 / $10 and $4 / $15 Sol rates, which gives a second independent confirmation of the headline price.

Sol vs Astra

Astra is the flagship. These numbers come from AlphaCorp unless noted, and the benchmark scores are at maximum reasoning effort.

GPT-6.2 SolGPT-6.2 Astra
Input / output per 1M tokens$2 / $10$8 / $40
Context window1,050,000 tokens1,050,000 tokens
Max output128K tokens128K tokens
Artificial Analysis Intelligence Index5762 (eesel AI)
FrontierCode 2.043.8%44.2%
Agents' Last Exam v1.173.0%77.6%
Terminal-Bench-Science 0.168.0%77.1%

The pattern is clear: on coding the two are close, while on science and long-horizon professional tasks Astra pulls ahead. At low reasoning effort the gap widens, so do not judge Sol by its cheapest setting.

How it compares with the competition

Artificial Analysis lists Sol at 57 on its Intelligence Index at maximum effort, with output of about 77 tokens per second. VentureBeat's table puts Google's new Gemini 4 Neon at 61 on the same index and Claude Sonnet 5.5 at 55. Sol is also the cheaper per-token option than Neon, which costs $3 / $15. Remember that these are different test conditions and vendor-chosen comparisons, so run your own tasks.

Who should use which

  • Sol: everyday coding, agent loops and routine work where cost per task matters. It is the better default for most people.
  • Astra: science, legal and other hard multi-step work where the extra accuracy pays for itself.
  • GPT-6.1 Sol users: same price and a two-point higher index score, but eesel AI reports a drop on one coding index, so test coding workloads before switching.

Caveats and what is unconfirmed

I could not open OpenAI's own announcement or pricing page (both returned errors), so the specs above come from third-party write-ups that cite them plus Artificial Analysis. Some sources differ on context details: one lists 1M tokens while others list 1.05M, and eesel AI notes a 922,000-token maximum input. Reported cost-per-task figures also differ between sources, so I have left them out. eesel AI's author sells an AI helpdesk product, which is worth keeping in mind.

Checked as of November 18, 2026. Independent explainer, not affiliated with OpenAI.

Sources: eesel AI, AlphaCorp, Artificial Analysis, VentureBeat.

Comments

Popular posts from this blog

Gemini 4 Neon vs Argon: What's Different and Which to Use

Game Awards Nominees 2026: Full List, Game of the Year Nominees and How to Vote

How to Find Trout Stocking Schedules Near You (Fall/Winter 2026)