On Tuesday, November 17, 2026, OpenAI launched two models at once: flagship GPT-6.2 Astra and everyday-tier GPT-6.2 Sol. On launch day, the independent board Artificial Analysis put Astra in the top 3 — 62 points at the Max tier, top 3 of 91 models, behind Claude Fable 5.2's 65; on AA-3D, Astra leads with 1,318 Elo. This article consolidates everything confirmed on launch day: the positioning, the pricing, the Sol relationship and the system-card thresholds.
GPT-6.2 Astra is OpenAI's flagship reasoning model — "Our most capable model for your hardest projects, from start to finish," in the model page's own words. Pricing is $8/$40 per 1M tokens (input/output), 20% below GPT-6 Astra's $10/$50, with cached input at $0.40 (down 60% from $1.00); context is 1.05M tokens (max input 922K), max output 128K; four effort tiers — Max / High / Medium / Low — plus an Ultrafast decoding tier; knowledge cutoff is May 18, 2026 (OpenAI model page, read directly by this site on Nov 18). Model ID gpt-6.2-astra, live the same day in ChatGPT Work, Codex and the API for Plus, Pro, Enterprise and Edu — not open to free users.
GPT-6.2 is a point release, not a new generation: GPT-6.1 Terra had only filled the gap between Sol and Luna a month earlier. Don't present it as a new generation. OpenAI has not disclosed the parameter count; the weights are closed, with no public release.
On launch day, Artificial Analysis scored it 62 (Max tier), top 3 of 91 models; AA-3D put it first at 1,318 Elo, and AnalystAgent at 62.5%. But Astra's real selling point isn't the score — it's the bill: $4.00 per index task, about 1/6.6 of Fable 5.2 ($8/$40 list price, $26.54 per task). Same list price, a 6.6× bill gap — the difference is token usage and task time, not the rate card. Astra finishes an index task in 11.5 minutes; Fable Max takes 45.3.
But the cheapness needs a discount of its own: Astra generated 65.8M tokens on the index task, more than GPT-6 Astra's 54.3M — the per-task cost fell because rates fell, not because the model got thriftier. See our pricing guide.
GPT-6.2 Sol, launched the same day, is the everyday tier at $2/$10 (a quarter of Astra), with an AA index of 57 — No. 11 of 91, up 2 points from GPT-6.1 Sol's 55 — at $1.01 per task, 7.9 minutes and 85 tok/s. OpenAI's own claim puts Sol within 0.5 points of Astra on Cognition FrontierCode 2.0 at about a fifth of the cost — vendor-reported, with no independent reproduction yet, but even discounted, Sol's positioning is clear: everyday work starts on Sol, and only the hardest tasks step up to Astra. Sol has its own topic queued; this article covers Astra only.
Ultrafast decoding launched alongside: 3× Standard rates through Feb 28, 2027, then 6× from Mar 1, 2027 (OpenAI model page). Sol's Ultrafast already reaches the $200/month Pro plan. On the system card, Astra's bio capability reaches the "High" threshold and cybersecurity "Critical"; OpenAI calls it the first released model at "High" for AI self-improvement — the safety headline of this launch, with a direct read of the full system card still pending.
Reference: OpenAI — GPT-6.2 Astra model page (read directly by this site, Nov 18, 2026) · Artificial Analysis — GPT-6.2 Astra model page (rewritten and organized by ModelBrief; facts per official sources and Artificial Analysis)