OpenAI Launches GPT-6.2 Astra and Sol on ChatGPT and API
11/18/2026 — 11/18, 05:17·10 sources (all)·21 reports (all)·In the 2026-11-18 briefing
Key takeaways
- OpenAI has released GPT-6.2 Astra and GPT-6.2 Sol, with both models now available through ChatGPT Work, Codex, and the API. GPT-6.2 Astra pr…
Background & analysis
In the early hours of November 18, 2026, Bloomberg Technology reported first that OpenAI had released a new version of its most advanced Astra model, roughly two months after the company scrapped an earlier update in order to put stronger safeguards in place around the technology. Shortly afterward, OpenAI's website went live with introduction pages for GPT-6.2 Astra and Sol, the link was picked up on the Hacker News front page, and users on V2EX relayed OpenAI's message on X confirming that both models were already available.
Additional outlets then added rollout details and benchmark figures. TestingCatalog reported that the models were being rolled out on ChatGPT Work, Codex, and the API, that Sol scored 57.3 on the Artificial Intelligence Index, and that Pro 200 users would get access to GPT-6.2 Sol Ultrafast. Artificial Analysis said GPT-6.2 Astra ranked second on its Intelligence Index and was priced at $8/$40 for input and output, 20% below GPT-6 Astra, with cached input dropping to $0.40 from $1.00. On evaluations, Astra led AA-3D with 1318 points, tops the research preview benchmark AA-Robotics v0.1, and scored 59% on the AA-Omniscience hallucination rate at max effort, compared with 55% for Sol. In the Coding Agent Index run with the Codex harness at max effort, Astra was flat versus GPT-6 Astra while Sol lost 3 points to GPT-6.1 Sol, driven mainly by a 14-point drop on SWE-Atlas-QnA-AA. Techmeme relayed OpenAI's positioning of Astra as its frontier model for complex professional work and Sol as its go-to model for everyday work. TestingCatalog cited claims that Astra beat GPT-6 Astra by 5.2 points on Agents' Last Exam at less than a third of the cost, and Claude Fable 5.2 by 5.5 points at just 5% of the cost. Artificial Analysis also reported that OpenAI led Harvey LAB-AA v1.1 through Astra for Law 6.2, a specialized configuration pairing GPT-6.2 Astra with an expanded legal search index. The event now stands at the stage of third-party disclosure of launch details and benchmark results, with OpenAI not yet providing further information on model parameters or full availability.
AI-generated from 14 reports · updated 3 hours ago
Latest turnOpenAI has released two new models, GPT-6.2 Astra and GPT-6.2 Sol, claiming clear gains in coding, research, professional office work and computer operation. The company also says the models complete more work with fewer tokens, so users get more done within existing usage limits. The reported improvements come from OpenAI's own claims, with no benchmark figures included in the report.
Related AI tools
- Heat index
- 931
- All sources
- 10
- All reports
- 21
- First seen
- 4 hours ago
Reports on this story headlines open the original
OpenAI has released two new models, GPT-6.2 Astra and GPT-6.2 Sol, claiming clear gains in coding, research, professional office work and computer operation. The company also says the models complete more work with fewer tokens, so users get more done within existing usage limits. The reported improvements come from OpenAI's own claims, with no benchmark figures included in the report.
cnBetaAI score 74OpenAI has released GPT-6.2 Astra and GPT-6.2 Sol, which are more capable and cheaper on the API than their predecessors, The Decoder reports. On the independent Artificial Analysis Intelligence Index, Astra ranks second behind Claude Fable 5.2. The report adds that Astra is the first released model to reach the High threshold for AI self-improvement on OpenAI's own scale, a level OpenAI classifies as a risk.
The DecoderAI score 76OpenAI has released GPT-6.2 Astra, a flagship model it says scores 5.5 points above Claude Fable 5.2 on a multi-step professional-work benchmark at 5% of the cost, alongside GPT-6.2 Sol for everyday tasks. According to Wallstreetcn, the launch comes nearly two months after GPT-6.1 Astra was held back over scope- and permission-boundary concerns, and just before Anthropic's hotly anticipated IPO.
华尔街见闻AI score 72OpenAI has launched GPT-6.2 Astra and Sol across ChatGPT Work, Codex and the API, according to TestingCatalog. Astra is said to outperform GPT-6 Astra at lower cost, with token pricing 20% cheaper.
TestingCatalogAI score 73Artificial Analysis' new AA-Robotics v0.1 benchmark puts GPT-6.2 Astra first at 79, 10 points above GPT-6 Astra and ahead of GPT-6.2 Sol (78), Claude Fable 5.2 (75) and Gemini 4 Hydrogen (58). Driving a real robot arm through Codex, it completed six of eight physical tasks, with the rest partly done.
X:Artificial Analysis (@artificialanlys)First-partyAI score 54GPT-6.2 Astra is priced at $8 per 1M input tokens and $40 per 1M output tokens, 20% below GPT-6 Astra, according to 9to5Mac. GPT-6.2 Sol keeps the $2 per 1M input and $10 per 1M output pricing of GPT-6.1 Sol.
TechmemeAI score 60TestingCatalog reports on X that OpenAI has officially announced GPT-6.2 Astra and Sol. The quoted figures claim GPT-6.2 Astra scores 5.2 points above GPT-6 Astra on the Agents' Last Exam benchmark at less than a third of the cost, and 5.5 points ahead of Claude Fable 5.2 at just 5% of the cost.
X:TestingCatalog (@testingcatalog)AI score 71Artificial Analysis reports that OpenAI leads Harvey's LAB-AA v1.1 legal benchmark with Astra for Law 6.2. Per its description, the configuration pairs GPT-6.2 Astra with an expanded legal search index, making it a specialized setup aimed at legal work rather than a standalone new model.
X:Artificial Analysis (@artificialanlys)First-partyAI score 41OpenAI released two models: GPT-6.2 Astra, pitched as its frontier model for the hardest projects and "complex professional work", and GPT-6.2 Sol, described as its efficient, versatile go-to model for everyday work. Techmeme relayed the launch from OpenAI's announcement.
TechmemeAI score 81Artificial Analysis measured hallucination rates of 59% for GPT-6.2 Astra and 55% for GPT-6.2 Sol at max effort on its AA-Omniscience metric. The metric tracks how often a model guesses when it does not know the answer to factual questions.
X:Artificial Analysis (@artificialanlys)First-partyAI score 50Artificial Analysis released AA-Robotics v0.1, a research-preview benchmark measuring how well language models understand, plan and act in the real world. GPT-6.2 currently leads the ranking.
X:Artificial Analysis (@artificialanlys)First-partyAI score 42GPT-6.2 Astra leads the AA-3D benchmark at 1318, ahead of Claude Fable 5.2 at 1271 and GPT-6.2 Sol at 1259. The benchmark, scored by human graders, measures how well models build interactive 3D experiences.
X:Artificial Analysis (@artificialanlys)First-partyAI score 40On the Artificial Analysis Coding Agent Index (Codex harness, max effort), GPT-6.2 Astra is flat versus GPT-6 Astra while GPT-6.2 Sol drops 3 points against GPT-6.1 Sol. The decline is driven by a 14-point fall on SWE-Atlas-QnA-AA.
X:Artificial Analysis (@artificialanlys)First-partyAI score 48A V2EX post citing an OpenAI post on X says GPT-6.2 Astra and GPT-6.2 Sol are now available. The post only links to the announcement and gives no model specs, pricing, or availability details, so the claims could not be independently verified from the supplied material.
V2EXAI score 53A V2EX post points to an OpenAI official page announcing GPT-6.2 Astra and GPT-6.2 Sol, confirming the release but little else. The post itself offers no details on capabilities, pricing, or how the models are being made available.
V2EXAI score 55Artificial Analysis reports that GPT-6.2 Astra is flat versus GPT-6 Astra on its Coding Agent Index, while GPT-6.2 Sol scores 3 points lower than GPT-6.1 Sol. The result was measured with the Codex harness at max effort and covers coding-agent performance only, not all tasks.
X:Artificial Analysis (@artificialanlys)First-partyAI score 54Artificial Analysis says OpenAI's GPT-6.2 Astra ranks #2 on its Intelligence Index and covers most of the Cost per Task Pareto frontier. Astra is priced at $8/$40, 20% below GPT-6 Astra, with cached input falling from $1.00 to $0.40. Sol is also part of the release.
X:Artificial Analysis (@artificialanlys)First-partyAI score 74According to TestingCatalog on X, OpenAI is rolling out GPT-6.2 Astra and GPT-6.2 Sol across ChatGPT Work, Codex, and its APIs. GPT-6.2 Sol scores 57.3 on the Artificial Intelligence Index, and Pro 200 users are getting access to a GPT-6.2 Sol Ultrafast tier.
OpenAI has published an official page introducing GPT-6.2 Astra and Sol. The link is currently on the Hacker News front page with 100 points and 23 comments.
Hacker News · 热榜AI score 75OpenAI has released a new version of Astra, its most advanced AI model, roughly two months after scrapping a prior update, Bloomberg reported. The earlier release was dropped to put stronger safeguards in place for the technology.
Bloomberg TechnologyAI score 73OpenAI released GPT-6.2 Astra and Sol, which the company says deliver better professional outputs, coding and computer use, plus lower cost, across ChatGPT Work, Codex and the API.
OpenAI NewsFirst-partyAI score 75
Other stories people are talking about
- Heat index 68Developer Compiles Roundup of 7 Email Marketing MCP Servers for Claude and ChatGPT1 sources
- Heat index 88NewGuide: use MCP to make AI assistants audit and fix AI-generated sites1 sources
- Heat index 134Claude Code Bypass Lets Agents Read Blocked .env Files2 sources
- Heat index 72Developer Shares Workflow for Dual-Agent Development with Claude Code and Codex1 sources
- Heat index 74OpenAI Report Shows Scientists Double Efficiency When Using ChatGPT1 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

