Ctrl K
Back to models

OpenAI

gpt-oss-20b

  • Released
  • Open weights
  • Apache 2.0
Compare
Intelligence Index

High setting

9.0#76 of 83measured models
Measured · Artificial Analysis v4.3.2
API price from
≈ $0.03 in / $0.14 out / 1M tokOpenRouter · DekaLLM · BF16
Context window
131.07K tok

Performance profile

ReasoningRank #79 of 107
68.8%
GPQA Diamond
CodingRank #68 of 94
38.9%
SciCode
Cost efficiencyRank #2 of 67
783.1pts / $
Intelligence ÷ eval cost

Capabilities

Context window

Modalities

Measured settings

Avg reasoning tokens / task — Artificial Analysis Intelligence Index v4.3.2: weighted average reasoning tokens per benchmark task, not a token limit. Source (opens in a new tab)

Integration

Reasoning trade-offs

Source (opens in a new tab)
Intelligence Index v4.3.2
gpt-oss-20b
Cost per task (USD) · log scale
gpt-oss-20bHigh
Intelligence Index
9.0
Cost / task
$0.0115
Evaluation time
1.3m

Pricing & access

API route
OpenRouter · DekaLLM · BF16API
Tier & contextStandard

USD per 1M tokens

Input
$0.029
Output
$0.14
Cached input
$0.029

Run locally

GGUFMXFP4
Context
Estimated memory

~13.7 GiB

RAM + VRAM

Estimate assumptions

One text chat at the selected context, with a 16-bit cache and app allowance. Operating system and image processing are extra; optimized caches can use less.

24 GiB RAMwithout GPU offload

Memory budgetOne text chat
Weights
11.3 GiB
Context cache
0.4 GiB16-bit cache
App allowance
2 GiB
Download
12.1 GB
Disk space
~19 GB
Available software
Ollama (opens in a new tab)LM Studio (opens in a new tab)Jan (opens in a new tab)KoboldCpp (opens in a new tab)llama.cpp (opens in a new tab)
Get GGUF (opens in a new tab)
Set up in Ollama · MXFP4 · 8K context

Save the following as Modelfile. The context matches the estimate above.

FROM hf.co/lmstudio-community/gpt-oss-20b-GGUF:MXFP4
PARAMETER num_ctx 8192

Then run these commands in your terminal:

ollama create llmcurrent-local -f Modelfile
ollama run llmcurrent-local

Ollama import instructions (opens in a new tab) · For LM Studio or Jan, open the app’s model search, enter lmstudio-community/gpt-oss-20b-GGUF, choose MXFP4, and set context to 8,192 tokens in the app’s model settings. Software links above open each app’s setup guide.

vLLM setup (opens in a new tab)

Benchmarks

Full leaderboardSource (opens in a new tab)

Humanity’s Last Exam

Frontier reasoning

gpt-oss-20b · High#88 of 126

  1. 1Claude Opus 5.5Max61.4%
  2. 2Claude Fable 5.1Max59.1%
  3. 3Gemini 4 ArgonHigh57.1%
  5. 85NVIDIA Nemotron 3 Nano 30B A3BReasoning11.4%
  6. 86Granite 4.2 30BReasoning11.2%
  7. 87North Mini CodeReasoning11.1%
  8. 88gpt-oss-20bHigh11.0%
  9. 89Hermes 4 - Llama-3.1 405BReasoning10.9%
  10. 90Nemotron 3.5 LightningReasoning10.6%
  11. 91Qwen3 Coder NextStandard10.1%

Peers use their best measured result on this benchmark.

Alternatives

Each model at its highest-scoring measured setting · Intelligence Index v4.3.2 and its evaluation cost per task.