DataLearner logo

GPT-6 SolvsGPT-6 Astra

Across 3 shared benchmarks, GPT-6 Astra leads overall: GPT-6 Sol wins 0, GPT-6 Astra wins 3, with 0 ties and an average score difference of -5.03.

OpenAI
GPT-6 Sol

OpenAI · 2026-09-22 · Reasoning model

OpenAI
GPT-6 Astra

OpenAI · 2026-09-03 · Reasoning model

GPT-6 Sol0 wins(0%)(100%)3 winsGPT-6 Astra

Benchmark scores

Grouped by capability, sorted by largest gap within each. 3 shared benchmarks.

Agent Level Benchmark

GPT-6 Astra 1/1
BenchmarkGPT-6 SolGPT-6 AstraDiff
Agents' Last Exam56.402 / 24Max (With Tools)59.301 / 24Max (With Tools)-2.90

AI Agent - Tool Usage

GPT-6 Astra 1/1
BenchmarkGPT-6 SolGPT-6 AstraDiff
OSWorld 2.064.407 / 13Max (With Tools)72.603 / 13Max (With Tools)-8.20

Coding and Software Engineer

GPT-6 Astra 1/1
BenchmarkGPT-6 SolGPT-6 AstraDiff
FrontierCode 1.1 Main49.307 / 9Max (With Tools)53.304 / 9Max (With Tools)-4

Specs

FieldGPT-6 SolGPT-6 Astra
PublisherOpenAIOpenAI
Release date2026-09-222026-09-03
Model typeReasoning modelReasoning model
ArchitectureDenseDense
ParametersNot availableNot available
Context length1.05M1.05M
Max output128K128K

API pricing

Prices use DataLearner records when available; missing fields are not inferred.

ItemGPT-6 SolGPT-6 Astra
Text input$2 / 1M tokens$10 / 1M tokens
Text output$10 / 1M tokens$50 / 1M tokens
Cache read$0.2 / 1M tokens$1 / 1M tokens
Cache write$2.5 / 1M tokens$12.5 / 1M tokens

Summary

  • GPT-6 Astraleads in:Agent Level Benchmark (1/1), AI Agent - Tool Usage (1/1), Coding and Software Engineer (1/1)

On average across the 3 shared benchmarks, GPT-6 Astra scores 5.03 higher.

Largest single-benchmark gap: OSWorld 2.0 — GPT-6 Sol 64.40 vs GPT-6 Astra 72.60 (-8.20).

Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.