OpenAI Launches GPT-6.2 Astra and Sol: Twin Models Split Reasoning and Creativity

TWIN AI BREAKTHROUGH

OpenAI has released GPT-6.2 Astra and GPT-6.2 Sol, two new models that improve professional outputs, coding, and computer-use capabilities while lowering API costs. The launch extends Ultrafast mode to more subscribers and introduces stricter cybersecurity safeguards alongside a new researcher access program.

Listen to today’s briefing (podcast)

Key takeaways

  • OpenAI launches GPT-6.2 Astra and Sol with 20% API price cut for Astra and new 3D computer-use capabilities.
  • GPT-6.2 Astra leads benchmarks in legal, science, and 3D tasks at a fraction of competitor cost.
  • Both models hit Critical cybersecurity threshold; OpenAI adds safeguards and plans Daybreak access for vetted researchers.
  • GPT-6.2 Sol matches GPT-6 Astra intelligence at roughly 15% cost per task, per Artificial Analysis.
  • Coding performance flatlines: Astra unchanged on Coding Agent Index, Sol drops three points on SWE-Atlas-QnA-AA.

OpenAI Launches GPT-6.2 Astra and Sol Models

OpenAI released GPT-6.2 Astra for demanding professional work and GPT-6.2 Sol for everyday use, available now to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, with API access for developers according to The Decoder. Pro-500 subscribers get Ultrafast mode on both models, while Pro-200 subscribers gain Ultrafast access for the first time with GPT-6.2 Sol. API pricing for Astra drops 20 percent to $8 per million input tokens, $0.40 for cached inputs, and $40 per million output tokens; Sol retains GPT-6.1 Sol pricing at $2, $0.10, and $10 per million tokens respectively.

On benchmarks, GPT-6.2 Astra scores 5.2 points above GPT-6 Astra on Agents’ Last Exam at less than a third of the cost and leads Anthropic’s Claude Fable 5.2 by 5.5 points at just 5 percent of the cost, with only Google’s Gemini 4 Carbon keeping pace per The Decoder. It more than doubles Claude Fable 5.2’s pass rate on the Harvey LAB-AA legal benchmark at about one-sixth the cost, gains 6.2 percentage points on the internal GDPval-v2 benchmark while using 22 percent fewer output tokens, and takes first on the AA-3D benchmark with 1,318 Elo points. GPT-6.2 Sol trails Astra by only 0.5 points on FrontierCode 2.0 at one-fifth the cost.

Both models reach the Critical threshold for cybersecurity in OpenAI’s Preparedness Framework, finding and exploiting vulnerabilities more effectively while writing more secure code per The Decoder. OpenAI added stricter safeguards for risky cybersecurity requests and plans to offer new models through OpenAI Daybreak for vetted security researchers with fewer guardrails; published benchmark numbers come from internal tests with safeguards and classifiers turned off. In the Artificial Analysis Intelligence Index, Astra finishes second behind Claude Fable 5.2 (up five points over GPT-6 Astra), while Sol reaches 57, matching GPT-6 Astra at about 15 percent cost per task; OpenAI owns the Pareto frontier on cost versus performance. Coding performance flatlined: Astra stays level on the Coding Agent Index, while Sol drops three points from GPT-6.1 Sol due to a 14-point slide on SWE-Atlas-QnA-AA.

Computer-use capabilities expanded: from two video frames, GPT-6.2 Astra reconstructed an entire set in Blender to pixel-perfect match including camera blur and professional lighting, then iterated on the environment, turned it into a walkable scene, and created a virtual background webcam app using the Decisions API to live-match webcam lighting per YouTube: Building 3D video sets with GPT-6.2 Astra. A separate demo shows Astra handling office tasks such as updating payables from email invoices, building a workbook for a one-year delay in savings on the Bremir integration, tracking net debt through 2031, showing next year’s EPS against plan, and creating presentation slides with tables per YouTube: GPT-6.2 Astra and GPT-6.2 Sol—Models built for the real world. The Hacker News discussion garnered 822 points and 503 comments per Hacker News.

Why it matters: The 20% API price cut, higher benchmark scores, and new 3D/computer-use capabilities mean cheaper, more powerful tools for building agents, automating workflows, and prototyping in code and Blender.

Tools mentioned in this briefing

  • n8n — n8n lets developers visually build AI agent workflows that can integrate OpenAI’s new models via API. Try n8n to orchestrate GPT-6.2 Astra and Sol in automated workflows without managing infrastructure. Try n8n
  • AI Engineering Book on Amazon — Developers exploring the new GPT-6.2 models may want a practical guide to building applications with large language models. See current options and reviews on Amazon. View on Amazon

Affiliate disclosure: some links in this description are affiliate links. If you sign up through them, SysArticles may earn a commission at no extra cost to you. We only mention tools that are relevant to the story. As an Amazon Associate I earn from qualifying purchases.

OpenAI’s latest models push the cost-performance frontier forward while tightening cybersecurity controls, giving IT teams more capable agents at lower operational expense.

Also: Türkçe podcast.

This briefing is produced with AI assistance from the linked primary sources. Narration uses a synthetic AI news-anchor voice, built with Higgs Materials licensed from Boson AI USA, Inc.

Liked this briefing? Get it in your inbox

Every morning: the stories that make tech leaders act, with official clips and primary sources. Free, 5-minute read.

Free. One click to unsubscribe. We never share your email. Privacy

OpenAI Launches GPT-6.2 Astra and Sol: Twin Models Split Reasoning and Creativity

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Scroll to top