AboutAI ToolsCursorClaude CodeLovableBolt.newAI tool reviewsTutorialsVibe codingAI agentsMobile apps with AIAll tutorialsBlogSponsorWork With MeJoin 285K+ Subscribers

18 November 2026 · 4 min read

OpenAI Launches GPT-6.2 Astra and Sol Models With 20% Astra API Price Cut

OpenAI released GPT-6.2 Astra and Sol on 17 November 2026, with a 20% price cut for Astra’s API and availability in ChatGPT, Codex, and GitHub Copilot. Early testers reported strong code review but struggled with long-horizon tasks.

On 17 November 2026, OpenAI released GPT-6.2 Astra and GPT-6.2 Sol, two closed models aimed at professional and agentic work [1][3][4]. Astra is the frontier tier for complex, long-horizon tasks, while Sol is the efficient everyday tier. Both are available through the OpenAI API as gpt-6.2-astra and gpt-6.2-sol, in ChatGPT Work and Codex, and in GitHub Copilot [1][2][4].

What was announced

GPT-6.2 Astra succeeds GPT-6 Astra and is designed for the hardest multistep work: building complex applications, producing documents that follow templates, and handling large codebase changes [1][4]. GPT-6.2 Sol replaces GPT-6.1 Sol for routine coding, writing, and multi-step workflows [1].

Both models share a 1,050,000-token context window (922,000 max input, 128,000 max output) and accept text and image inputs [1]. Astra has a knowledge cutoff of 18 May 2026; Sol’s is 1 June 2026. Neither model accepts reasoning effort set to none. OpenAI recommends keeping prompt prefixes unchanged and appending new instructions to improve cache reuse and preserve reasoning across turns [1].

Pricing and availability

Astra’s standard API prices are 20% lower than GPT-6 Astra: $8 per million input tokens and $40 per million output tokens. Cached input is $0.40 per million tokens, a 60% reduction [1]. Sol keeps GPT-6.1 Sol pricing at $2 per million input and $10 per million output, with cached input at $0.10 [1]. For requests over 272,000 input tokens, input and cache rates double and output costs 1.5×. Batch processing is half the standard rate. Ultrafast mode costs 3× standard through 28 February 2027, then 6× [1].

In ChatGPT Work and Codex, the models are available to Plus, Pro, Enterprise, and Edu users [4]. The $200/month Pro plan gets Ultrafast mode for Sol; the $500/month Pro plan gets Ultrafast for both models. Eligible Enterprise and Edu workspaces can enable Ultrafast via an admin setting [4].

GitHub Copilot Pro+, Max, Business, and Enterprise users can select the models in VS Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, Copilot CLI, the coding agent, the Copilot app, github.com, and GitHub Mobile. Rollout is gradual [2].

Performance and benchmarks

On the Artificial Analysis Intelligence Index, Astra scores 62 at maximum effort, second behind Claude Fable 5.2 (65). Sol scores 57, matching GPT-6 Astra’s score at roughly 15% of its cost per task [1]. Astra ranks first on AA-3D with 1,318 Elo, ahead of Fable 5.2 (1,271). However, Astra’s hallucination rate on AA-Omniscience is 59%; Sol’s is 55% [1]. Sol’s Coding Agent Index dropped 3 points compared to its predecessor [1].

OpenAI’s internal benchmarks show cost-adjusted gains. On Agents’ Last Exam v1.1, Astra scores 5.2 points above GPT-6 Astra at under a third of the cost, and leads Fable 5.2 by 5.5 points at 5% of the cost. On Terminal-Bench-Science 0.1, Astra gains 9 points over GPT-6 Astra at under a third of the cost. On GDPval v2, Astra gains 6.2 points with 22% fewer output tokens; Sol gains 5.3 points over GPT-6.1 Sol. On FrontierCode 2.0, Sol lands within 0.5 points of Astra at about one-fifth the cost [1].

In GitHub Copilot, early testing showed both models caught significantly more real issues in code review than earlier GPT-6 models. Sol caught nearly twice as many real issues as GPT-6.1 Sol while keeping findings accurate [2].

Early reception

Several AI power users who tested the models before launch reported mixed experiences. Investor Matt Shumer said the models were “essentially unusable” [3]. Developer Theo Browne called 6.2 Astra “the worst model OpenAI has ever released,” citing difficulties with long-running projects and a tendency to produce similar outputs across different tasks [3]. Wharton professor Ethan Mollick found Astra “very uneven & weird” and harder to steer than rivals, though it built a playable space elevator in Blender and Unity [3].

Dan Shipper, CEO of Every, said Astra performed similarly to its predecessor on most coding benchmarks but often stopped early on end-to-end builds. Its writing could be hard to follow, and its design outputs had a “noticeable house style” that was difficult to vary [3]. Browne, however, credited Astra with catching issues other models missed in code reviews and helping him devise a reverse proxy expected to save $25,000 per month [3].

Some users praised Astra’s strengths in animated visuals, motion graphics, VFX, and agentic video editing [3]. OpenAI says many of its own employees already use Sol as their default for technical work [3].

What it means for developers

The 20% price cut on Astra and Sol’s unchanged low pricing make frontier-level reasoning more accessible. Sol’s near-Astra performance on FrontierCode 2.0 at one-fifth the cost offers a compelling option for coding tasks. GitHub Copilot integration brings improved code review capabilities, with both models catching more real issues while using fewer tokens [2].

However, the high hallucination rates and reports of unreliable behaviour on long-horizon agentic tasks suggest developers should evaluate carefully before delegating complex, multi-step work. Astra’s strong 3D and game development results may benefit specialised workflows, but its inconsistent steering and tendency toward repetitive outputs could require additional oversight.

Sources

  1. OpenAI Releases GPT-6.2 Astra and Sol With a 20% Cheaper Astra API
  2. OpenAI's GPT-6.2 Astra and GPT-6.2 Sol now available - GitHub Changelog
  3. 'Essentially unusable': OpenAI launches GPT-6.2 Astra and Sol to negative reception from AI power users
  4. OpenAI upgrades ChatGPT and Codex with new GPT-6.2 Astra and Sol models
  • gpt-6.2
  • openai
  • astra
  • sol
  • api

Researched and drafted with AI from the sources above.

Learn to build with AI

Hands-on tutorials on Cursor, Claude Code, AI agents and more.