Back

OpenAI

GPT-5.6 Luna

ReasoningClosed weightsModels.dev
Intelligence
37.3
Rank 83 of 744 measured models
Coding
71.4
Rank 50 of 298 measured models
Output speed
115 t/s
Rank 142 of 403 measured models
Price/1M
$0.45
Rank 256 of 659 measured models
Context window
1M
Rank 84 of 1199 measured models
Cost per task
$0.178
Rank 68 of 193 measured models

GPT-5.6 Luna (Max) scores 37.3 on the Artificial Analysis Intelligence Index, ranking 83rd of 744 measured models. The closest models on this index are Claude Sonnet 5 (38.2), GPT-6 Luna (38.1), and MiMo-V2.6-Flash (37.9). At $0.45 per million tokens (3:1 input/output blended price), it is cheaper than Claude Sonnet 5 ($4), more expensive than GPT-6 Luna ($0.20) and MiMo-V2.6-Flash ($0.175). With an output speed of 115 t/s, it is faster than Claude Sonnet 5 (74.9 t/s) and MiMo-V2.6-Flash (53.8 t/s), about as fast as GPT-6 Luna (116 t/s).

Artificial Analysis Indices
Maximum
Intelligence37.3
Rank 83 of 744 measured models
Coding71.4
Rank 50 of 298 measured models
Agentic42.1
Rank 37 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 & Accounting40.1
Rank 76 of 218 measured models
Strategy & Ops44.1
Rank 68 of 218 measured models
Legal36.9
Rank 95 of 218 measured models
Healthcare & Medical35.7
Rank 41 of 82 measured models
Engineering36.2
Rank 83 of 217 measured models
Economics42.7
Rank 97 of 225 measured models
Standard Benchmarks
GPQA Diamond91.1%
Rank 52 of 619 measured models
HLE39.5%
Rank 87 of 664 measured models
SciCode53.6%
Rank 73 of 231 measured models
AA-LCR83.7%
Rank 15 of 575 measured models
Terminal-Bench 2.180.9%
Rank 43 of 244 measured models
Terminal-Bench v411.6%
Rank 84 of 218 measured models
τ-Bench Banking31.1%
Rank 65 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