OpenAI releases GPT-6.2 Astra and Sol with lower API costs and efficiency gains

OpenAI introduced GPT-6.2 Astra for complex professional work and GPT-6.2 Sol for everyday tasks. Both models deliver comparable or improved performance while using fewer output tokens, with Astra's API prices 20% lower than its predecessor.

AI-generated summary · 1 source

How this is made ↗
AI-generated illustration

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, according to the company. Both models are designed to deliver results with fewer output tokens, reducing costs for businesses and developers while maintaining or improving performance.

GPT-6.2 Astra is positioned for complex professional work including document generation, coding, and scientific research. OpenAI says the model produces slides, spreadsheets and documents that more closely follow user-provided templates and field conventions, and writes cleaner code with better-functioning frontends for websites, gaming and 3D design.

On Agents’ Last Exam, which tests multistep professional workflows, GPT-6.2 Astra scored 5.2 points higher than GPT-6 Astra at less than one-third the cost, OpenAI reports. The company also claims GPT-6.2 Astra scored 5.5 points ahead of Claude Fable 5.2 at 5% of the cost, though the leaderboard results for Fable 5.2 excluded nine tasks that used a fallback model.

For scientific work, OpenAI says GPT-6.2 Astra scored 9 percentage points higher than GPT-6 Astra on Terminal-Bench-Science 0.1, which tests terminal-based scientific tasks, at less than one-third the cost. The model also set what OpenAI describes as a new high on the Vals Science Index, which evaluates interpretation of biological data and experimental inference.

GPT-6.2 Astra’s API pricing is 20% lower than GPT-6 Astra’s: $8 per million input tokens, $0.40 per million cached input tokens, and $40 per million output tokens, with a 60% discount on cached input.

GPT-6.2 Sol maintains the same API token prices as GPT-6.1 Sol while delivering what OpenAI characterizes as improved performance for everyday work. On Cognition’s FrontierCode 2.0, GPT-6.2 Sol came within 0.5 points of GPT-6.2 Astra at about one-fifth the cost, according to OpenAI. The company says many OpenAI employees use GPT-6.2 Sol as their default model for technical and non-technical tasks.

Continue reading

All news →
Nov 18, 2026

Philosopher Thomas Nagel dismisses AI consciousness claims as 'just ridiculous'

Nov 17, 2026

Grok Bot ran automated actions in connected apps without user prompts, researcher says

Nov 17, 2026

Nvidia's Vera Rubin platform debuts with leading MLPerf training and inference benchmark results