modelgrep
O

OpenAI: gpt-oss-20b

openai/gpt-oss-20b

103rd smartest of 178Cheaper than 98% of paidReasoningToolsJSON
Use via OpenRouter ↗
Intelligence
24.5
103rd of 178
Design Elo
977
Data Viz
Speed
391
7th fastest
Latency
247ms
first token
Input price
$0.029
6th cheapest
Context
131K

How it compares

Smarter than42%
of all ranked models
Faster than98%
of all ranked models
Cheaper than98%
of all ranked models

Overview

gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for...

Benchmarks

independent · via OpenRouter
Artificial Analysis42th percentile
Intelligence Index
24.5
Coding Index
18.5
Agentic Index
27.6
GPQA Diamond
69%
Humanity's Last Exam
10%
SciCode
34%
Tau²-Bench (agentic)
60%
Design Arena · Elo471 tournaments
Data Viz
977
Website
897

Providers & pricing (14)

ProviderIn $/MOut $/MUptime
DekaLLMbf16$0.029$0.14098.3%
WandBfp4$0.030$0.13099.2%
DeepInfrabf16$0.030$0.14099.7%
Phala$0.040$0.15065.2%
Novitafp4$0.040$0.15099.8%
SiliconFlowfp8$0.040$0.180100%
Parasailfp4$0.040$0.20099.5%
Together$0.050$0.200
Amazon Bedrock$0.070$0.150
Amazon Bedrock$0.070$0.15098.8%
Google$0.070$0.25099.7%
Fireworks$0.070$0.30098.7%
Groq$0.075$0.30099.9%
NextBitfp8$0.100$0.450100%

Specifications

Context window131K
Max output
Knowledge cutoffJun 2024
Input modalitiestext
Output modalitiestext
Prompt caching
Cache read price
ModeratedNo

gpt-oss-20b FAQ

How much does gpt-oss-20b cost?

gpt-oss-20b costs $0.029 per million input tokens and $0.140 per million output tokens via OpenRouter, making it 6th cheapest of 298 paid models.

How smart is gpt-oss-20b?

gpt-oss-20b scores 24.5 on the Artificial Analysis Intelligence Index, ranking 103rd of 178 benchmarked models, with a GPQA Diamond score of 69%.

How fast is gpt-oss-20b?

gpt-oss-20b generates around 391 tokens per second with 247ms time-to-first-token (p50), the 7th fastest tracked model.

What is gpt-oss-20b's context window?

gpt-oss-20b supports a 131K-token context window. It accepts text input.

Compare head-to-head