Back

OpenAI

gpt-oss-20b

ReasoningOpen weightsModels.devHugging Face
Intelligence
9.0
Rank 462 of 744 measured models
Coding
20.7
Rank 225 of 298 measured models
Output speed
226 t/s
Rank 26 of 403 measured models
Price/1M
$0.098
Rank 87 of 659 measured models
Context window
131K
Rank 738 of 1199 measured models
Cost per task
$0.0125
Rank 16 of 193 measured models

gpt-oss-20b (High) scores 9.0 on the Artificial Analysis Intelligence Index, ranking 462nd of 744 measured models. The closest models on this index are Devstral Medium (9.0), GPT-4o (March 2025, chatgpt-4o-latest) (9.0), and Tri-21B-Think (9.0).

Artificial Analysis Indices
High
Intelligence9.0
Rank 462 of 744 measured models
Coding20.7
Rank 225 of 298 measured models
Math89.3
Rank 30 of 270 measured models
Agentic1.2
Rank 150 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 ↗Price / 1M tokens
Capability indexes
Performance in specific industries, measured by Artificial Analysis
Finance & Accounting7.3
Rank 203 of 218 measured models
Strategy & Ops4.8
Rank 201 of 218 measured models
Legal7.3
Rank 208 of 218 measured models
Healthcare & Medical6.6
Rank 82 of 82 measured models
Engineering11.7
Rank 195 of 217 measured models
Economics11.3
Rank 214 of 225 measured models
Standard Benchmarks
MMLU Pro74.8%
Rank 183 of 349 measured models
GPQA Diamond68.8%
Rank 333 of 619 measured models
HLE11.0%
Rank 313 of 664 measured models
LiveCodeBench77.7%
Rank 37 of 343 measured models
SciCode38.9%
Rank 182 of 231 measured models
AIME 202589.3%
Rank 30 of 270 measured models
IFBench65.1%
Rank 105 of 450 measured models
AA-LCR34.7%
Rank 390 of 575 measured models
Additional benchmarks2
Design Arena · Dataviz39.7%
Design Arena · Website27.9%
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