OpenAI

GPT-6.2 Astra

Intelligence
45.7
Rank 49 of 813 measured models
Output speed
43.4 t/s
Rank 427 of 476 measured models
Context window
1M
Rank 120 of 1420 measured models
Cost per task
$0.662
Rank 134 of 255 measured models

GPT-6.2 Astra (Low) is a text AI model from OpenAI, released on Nov 17, 2026, with a context window of 1M tokens. GPT-6.2 Astra (Low) scores 45.7 on the Artificial Analysis Intelligence Index, ranking 49th of 813 measured models. The closest models on this index are GPT-5.6 Sol (47.0), Kimi K3.1 Pro (46.5), and Claude Opus 5 (44.2). With an output speed of 43.4 t/s, it is faster than Kimi K3.1 Pro (31.2 t/s), slower than GPT-5.6 Sol (104 t/s) and Claude Opus 5 (80.6 t/s). Every benchmark side by side: GPT-6.2 Astra (Low) vs Kimi K3.1 Pro.

Artificial Analysis Indices

Low
Intelligence45.7
Rank 49 of 813 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 & Accounting45.1
Rank 74 of 283 measured models
Strategy & Ops48.6
Rank 78 of 283 measured models
Legal45.6
Rank 75 of 283 measured models
Healthcare & Medical41.9
Rank 69 of 210 measured models
Engineering50.5
Rank 49 of 281 measured models
Economics43.6
Rank 97 of 289 measured models

Cyber Index

Artificial Analysis · v1
Cyber Index33.7%Rank 27 of 36 measured models
Successes
33.7%
Safety blocks
33.3%
Failures
33.0%
CWE-Bench-AA61.7%
DeepsecBench-AA39.4%
CyberGym-E2E-AA0.0% (safety blocks)
Cost per task—

* The model or provider declined some tasks on safety grounds.

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