Open Brain
OpenAIOfficial ReleaseOfficial Source

Introducing GPT-6.2 Astra and Sol

GPT-6.2 Astra targets complex coding and multi-step work; Sol targets everyday agent workflows. Compare them on your codebase and task mix, since benchmark results and API costs vary by model and setup.

OpenAI · Nov 17, 2026
Open Source Open MarkdownOpen JSON
Source Summary

**GPT-6.2 Astra** is positioned for complex projects, including changes in existing codebases; **GPT-6.2 Sol** is for everyday coding and agent workflows. OpenAI reports Astra scored 6.2 percentage points higher than GPT-6 Astra on GDPval v2 while using 22% fewer output tokens, and Sol came within 0.5 points of Astra on FrontierCode 2.0 at about one-fifth the cost, both at max settings. For coding agents, try Astra on work that needs implementation and testing across an existing codebase, and Sol on routine tasks. Compare review effort and token use on your own tasks before changing defaults. These are vendor-reported evaluations, and results depend on the benchmark and configuration. OpenAI also says tighter cybersecurity safeguards may restrict some legitimate vulnerability research workflows.

Practical Implication

No separate implication has been recorded yet.

Agent-Ready Context
**GPT-6.2 Astra** is positioned for complex projects, including changes in existing codebases; **GPT-6.2 Sol** is for everyday coding and agent workflows. OpenAI reports Astra scored 6.2 percentage points higher than GPT-6 Astra on GDPval v2 while using 22% fewer output tokens, and Sol came within 0.5 points of Astra on FrontierCode 2.0 at about one-fifth the cost, both at max settings.
For coding agents, try Astra on work that needs implementation and testing across an existing codebase, and Sol on routine tasks. Compare review effort and token use on your own tasks before changing defaults.
These are vendor-reported evaluations, and results depend on the benchmark and configuration. OpenAI also says tighter cybersecurity safeguards may restrict some legitimate vulnerability research workflows.
Connected Context · Feed7 Judgment

This distinguishes Astra for complex codebase work from Sol for routine workflows and adds vendor-reported benchmark results and token-cost comparisons. The guidance remains workload-specific: the reported scores do not establish which model will perform best in a particular agent setup, and tighter cyber safeguards may constrain some research tasks.

Context Map
modelcodingresearch#coding-agents#model-selection#agent-reliability