Back

OpenAI

GPT-6 Astra

ReasoningClosed weights
Intelligence
50.9
Rank 17 of 744 measured models
Coding
77.1
Rank 8 of 298 measured models
Output speed
42.6 t/s
Rank 353 of 403 measured models
Price/1M
$20
Rank 638 of 659 measured models
Context window
1M
Rank 84 of 1199 measured models
Cost per task
$1.73
Rank 156 of 193 measured models

GPT-6 Astra (High) scores 50.9 on the Artificial Analysis Intelligence Index, ranking 17th of 744 measured models. The closest models on this index are GPT-6.1 Sol (51.8), Claude Opus 5 (50.8), and Claude Fable 5 (batch) (49.6). At $20 per million tokens (3:1 input/output blended price), it is more expensive than GPT-6.1 Sol ($4) and Claude Opus 5 ($10). With an output speed of 42.6 t/s, it is slower than GPT-6.1 Sol (47.9 t/s) and Claude Opus 5 (54.7 t/s).

Artificial Analysis Indices
High
Intelligence50.9
Rank 17 of 744 measured models
Coding77.1
Rank 8 of 298 measured models
Agentic48.2
Rank 25 of 189 measured models
Reasoning levels
Each point is a reasoning level of this model: more reasoning often raises the score but costs more. Select a point to open that level.
Intelligence Indexmost efficient ↗Cost per task
Capability indexes
Performance in specific industries, measured by Artificial Analysis
Finance & Accounting53.2
Rank 17 of 218 measured models
Strategy & Ops55.5
Rank 17 of 218 measured models
Legal57.9
Rank 11 of 218 measured models
Engineering53.0
Rank 18 of 217 measured models
Economics58.4
Rank 17 of 225 measured models
Standard Benchmarks
GPQA Diamond94.9%
Rank 4 of 619 measured models
HLE53.1%
Rank 17 of 664 measured models
SciCode55.4%
Rank 48 of 231 measured models
AA-LCR80.0%
Rank 77 of 575 measured models
Terminal-Bench 2.189.9%
Rank 3 of 244 measured models
Terminal-Bench v454.0%
Rank 12 of 218 measured models
τ-Bench Banking40.0%
Rank 32 of 211 measured models
Models at a similar level
The closest models on the Intelligence Index, one configuration per family.

Preferences and analytics

Rybbit: visits, performance, clicks, copies, forms and session replay (masked inputs), with your permission. Refusing does not limit the site.

Privacy

Allow navigation analytics?

With your permission, Rybbit measures visits, performance and interactions (clicks, copies, forms), including session replay, to improve BenchSift. Input values are masked in replays. Refusing does not limit the site.

Your choice is stored for 180 days. Change it at any time using “Privacy preferences” in the footer. Withdrawal reloads the page.

No choice stored: Rybbit is disabled.

Privacy