GPT-6 Astra
OpenAI · 5 thinking configurations
Intelligence45.8
Input / 1M$10.00
Output / 1M$50.00
Context1M
Tariff: OpenAI · Source: artificial-analysis · USD per million tokens
Thinking configuration
What does more thinking buy you?
Compare this model’s measured thinking configurations.
Scroll the chart ↔ · tap a point for details
Efficient frontier: best measured score at each cost5 measured configurations · 0 without both values
Selected: thinking low · Click a point to switch configuration
GPT-6 Astra is OpenAI's flagship model for demanding end-to-end work. It is suited for advanced analysis, software engineering, deep research, scientific work, and document creation, with particular strengths in long-horizon...
Intelligence IndexArtificial Analysis composite index, not a percentage. Different versions of the index are not directly comparable.
45.8 pointsDeepSWE · pass@1Software-engineering tasks on mini-swe-agent v1.1. Pass@1 across evaluated runs, with measured cost per task. The 95% intervals may overlap.
67.0 % ±1.3GPQA DiamondGraduate-level scientific questions. Measures specialist reasoning rather than web-search quality.
93.1 %Humanity’s Last ExamDifficult questions across many disciplines, compared under the same evaluation setup.
49.2 %SciCodeCode generation for scientific research problems.
54.1 %Terminal-Bench 2.1Agent tasks in a terminal environment. Results also depend on the evaluation harness.
88.0 %MMMU-ProMultimodal understanding of academic problems that include images.
84.6 %Omniscience accuracyFactual accuracy in Artificial Analysis's Omniscience evaluation.
59.5 %Specifications from the catalog
Inputfile, image, text
Outputtext
Tool callingYes
Structured outputYes
Speed47.2 tok/s
First token2.83 s
Maximum output128K
Catalog listing date2026-09-04
DeepSWE v1.1 · 113 tasks · mini-swe-agent · cost $1.60 / task · 20 mean steps · 10.6K mean output tokens. Published by the source 9/22/2026.