{"slug": "gpt-6-2-astra-and-sol-more-work-fewer-tokens-lower-cost", "title": "GPT-6.2 Astra and Sol: More Work, Fewer Tokens, Lower Cost", "summary": "OpenAI released GPT-6.2 Astra and GPT-6.2 Sol, two models the company says complete demanding professional tasks with fewer output tokens and lower cost than their predecessors. OpenAI reports Astra scores 5.2 points higher than GPT-6 Astra on Agents' Last Exam at less than a third of the cost and 5.5 points ahead of Claude Fable 5.2 at 5% of the cost, while Sol comes within 0.5 points of Astra on FrontierCode 2.0 at about one-fifth the cost. On GDPval v2, Astra improves 6.2 percentage points over GPT-6 Astra while using 22% fewer output tokens, and Sol improves 5.3 points over GPT-6.1 Sol.", "body_md": "**OpenAI just changed the math on frontier AI.** Not by making a bigger model, but by making a smarter, cheaper one that finishes real professional work with less waste.\n\nThe [launch](https://openai.com/index/introducing-gpt-6-2-astra-sol/) pairs two models with two very different jobs. Astra is for your hardest projects. Sol is for everything else. Both deliver better results on demanding tasks with fewer output tokens, so you can get more done within your usage limits.\n\n## Which One Should You Use?\n\n``` php\nflowchart TD\n    A[What are you doing today?] --> B{Is it high stakes, complex, or client facing?}\n    B -->|Yes| C[Use GPT-6.2 Astra]\n    B -->|No| D[Use GPT-6.2 Sol]\n    C --> E[Slides, memos, code, science, legal]\n    D --> F[Everyday work, drafts, planning, coding help]\n```\n\n## Astra Is Built for Work That Has to Be Right\n\nAstra is OpenAI's new frontier model for complex professional work. Its slides, spreadsheets, and documents follow your templates and your field's conventions more closely, with cleaner code and better frontends for websites, gaming, and 3D design.\n\nOn Agents' Last Exam, which tests multistep professional work, Astra scores 5.2 points higher than GPT-6 Astra at less than a third of the cost, and 5.5 points ahead of Claude Fable 5.2 at just 5% of the cost.\n\nFor scientists, it sets a new high on Terminal-Bench-Science 0.1, scoring 9 percentage points higher than GPT-6 Astra at less than one-third the cost, and also leads on the Vals Science Index for interpreting real biological data.\n\n## Sol Is the Everyday Powerhouse\n\nSol is smarter, more efficient, and lets you complete more tasks on the same paid plan. It powers through multistep agentic workflows, writes clearly and naturally, and is intuitive to collaborate with.\n\n- On FrontierCode 2.0, Sol comes within 0.5 points of Astra at about one-fifth the cost.\n- On [LifeVal](https://openai.com/index/lifeval/) , which measures challenging everyday tasks like shortlisting kindergartens after a move or planning meals for a parent adjusting to a new diagnosis, Sol outscores GPT-6 Astra at about one-seventh the cost.\n\nThat is why many OpenAI employees already use Sol as their default, for technical and non-technical tasks.\n\n## How the Efficiency Works\n\n``` php\nflowchart LR\n    A[Real task plus files] --> B[Astra or Sol]\n    B --> C[Fewer output tokens]\n    C --> D[Same quality or better]\n    D --> E[Lower cost plus faster turnaround]\n```\n\n## Professional Output That Needs Fewer Edits\n\nAstra was trained to create clearer, better-designed slides, spreadsheets, and documents that follow the conventions professionals use in law, sales, finance, and accounting, with more direct writing and less filler.\n\nOn Harvey LAB-AA v1.1, which tests legal work across source documents and detailed briefs, Astra more than doubles Claude Fable 5.2's pass rate at about one-sixth the cost.\n\nOn GDPval v2, drawn from real projects in consulting, finance, and design, Astra is up 6.2 percentage points from GPT-6 Astra while using 22% fewer output tokens. Sol improves 5.3 points from GPT-6.1 Sol.\n\n``` php\nflowchart TD\n    A[Source docs plus template] --> B[Find the relevant facts]\n    B --> C[Apply field conventions]\n    C --> D[Draft the deliverable]\n    D --> E[Check and refine]\n    E --> F[Ready to send deck, workbook, or memo]\n```\n\nFor slides, Astra matches layout, colors, and typography down to the exact font size, and carries the template's writing style into new slides. Early testers found the decks easier to use directly with fewer edits.\n\n## It Can Also Build and Design\n\nAstra brings more precise computer use for CAD and simulation, plus stronger visual reasoning for web design. Those gains carry over to 3D design, where Astra ranks first on AA-3D with 1,318 Elo, followed by Claude Fable 5.2 at 1,271 and Sol at 1,259.\n\n## Stronger, With Tighter Guardrails\n\nBoth models mark another jump in cyber capabilities, finding and exploiting vulnerabilities more effectively, writing more secure code, and producing better patches. Like GPT-6 Astra, they meet the Critical threshold for cybersecurity.\n\nBecause of that, they launch with tighter safeguards for higher-risk cybersecurity assistance. Approved defenders can access reduced refusals through [OpenAI Daybreak](https://openai.com/daybreak/), and OpenAI says it will work to bring GPT-6.2 models there.\n\nBoth models also meet the High threshold for biological and chemical capabilities under OpenAI's [Preparedness Framework](https://openai.com/index/updating-our-preparedness-framework/), with increased restrictions on dual-use requests. Astra is the first externally released model to meet the High threshold for AI Self-improvement, which OpenAI describes as equivalent to giving every researcher an additional, highly capable research engineer as an assistant.\n\nTo manage that risk, OpenAI strengthened containment in research environments and built Control Room, a system that tracks research workloads and can halt further model output when intervention is needed. See the full [GPT-6.2 system card](https://deploymentsafety.openai.com/gpt-6-2-astra-sol) for methodology and full results.\n\n## Pricing and Availability\n\nAstra and Sol are available to all Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, plus developers via the API.\n\n| Model | Input | Cached Input | Output | \n|---|---|---|---|\n| GPT-6.2 Astra | $8 | $0.40 | $40 | \n| GPT-6.2 Sol | $2 | $0.10 | $10 | \n\nPrices are per million tokens. Astra is 20% cheaper than GPT-6 Astra on input and output, with a 60% discount on cached input. Sol keeps GPT-6.1 Sol's rates.\n\n## The Takeaway\n\nGPT-6.2 is not about doing more for more. It is about doing better work for less: fewer tokens, fewer edits, fewer rounds of revision, and a lower bill at the end.\n\nIf the deliverable matters, start with Astra. If you want speed and value for everyday work, Sol is now close enough to frontier that for most tasks, you will not miss it.", "url": "https://wpnews.pro/news/gpt-6-2-astra-and-sol-more-work-fewer-tokens-lower-cost", "canonical_source": "https://zyvop.com/gpt-6-2-astra-and-sol-more-work-fewer-tokens-lower-cost-fplm1", "published_at": "2026-11-17 18:36:20+00:00", "updated_at": "2026-11-17 20:17:56.018312+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-agents", "ai-products"], "entities": ["OpenAI", "GPT-6.2 Astra", "GPT-6.2 Sol", "Claude Fable 5.2", "GPT-6 Astra", "GPT-6.1 Sol"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/gpt-6-2-astra-and-sol-more-work-fewer-tokens-lower-cost", "markdown": "https://wpnews.pro/news/gpt-6-2-astra-and-sol-more-work-fewer-tokens-lower-cost.md", "text": "https://wpnews.pro/news/gpt-6-2-astra-and-sol-more-work-fewer-tokens-lower-cost.txt", "jsonld": "https://wpnews.pro/news/gpt-6-2-astra-and-sol-more-work-fewer-tokens-lower-cost.jsonld"}}