OpenAI, open weights

# gpt-oss-20b

> gpt-oss-20b by OpenAI, released August 2025. Ranked #255 of 354 with a Noometry Index of 32.5. API: $0.018 in / $0.09 out per M tokens. 131K context. Scores, sources and comparisons.
- Canonical page: https://noometry.com/models/gpt-oss-20b
- Last updated: 2026-10-10
- Title: gpt-oss-20b Benchmarks, Price & Rank (October 2026)

gpt-oss-20b by OpenAI ranks 255th of 354 ranked models on the Noometry Index as of October 2026, with a score of 32.5. Its strongest category is math, where it ranks 103rd. API pricing starts at $0.018 per million input tokens and $0.09 per million output tokens, with a 131K-token context window.

Last verified October 10, 2026

## Specifications

- **Noometry rank:** #255 of 354
- **Index score:** 32.5
- **Evidence:** Confirmed 34 results
- **Provider:** [OpenAI](https://noometry.com/providers/openai)
- **Released:** August 5, 2025
- **Weights:** Open weights
- **Reasoning:** Yes
- **Context window:** 131K
- **Max output:** 16K
- **Input price:** $0.018 / M
- **Output price:** $0.09 / M
- **Blended price:** $0.036 / M
- **Output speed:** 96 tokens/s [Kagi](https://help.kagi.com/kagi/ai/llm-benchmark.html)
- **Value:** #1 of 219
- **Knowledge cutoff:** Unknown
- **Input:** text
- **Hugging Face:** [openai/gpt-oss-20b](https://huggingface.co/openai/gpt-oss-20b)

## Category scores

Each category score combines every public result we have in that category.

gpt-oss-20b category scores

1.  Coding 37.6
2.  Agentic & Tool Use 9.3
3.  Reasoning 19.3
4.  Math 39.4
5.  Knowledge 34.6
6.  Multilingual 42.2
7.  Instruction Following 61.8
8.  Long Context 37.9
9.  Writing & Preference 35.5
10.  020406080

gpt-oss-20b category ranks
| Category | Score | Rank | Results |
| --- | --- | --- | --- |
| [Coding](https://noometry.com/best/coding) | 37.6 | #192 | 3 |
| [Agentic & Tool Use](https://noometry.com/best/agentic) | 9.3 | #154 | 1 |
| [Reasoning](https://noometry.com/best/reasoning) | 19.3 | #261 | 6 |
| [Math](https://noometry.com/best/math) | 39.4 | #103 | 3 |
| [Knowledge](https://noometry.com/best/knowledge) | 34.6 | #195 | 4 |
| [Multilingual](https://noometry.com/best/multilingual) | 42.2 | #197 | 1 |
| [Instruction Following](https://noometry.com/best/instruction-following) | 61.8 | #240 | 2 |
| [Long Context](https://noometry.com/best/long-context) | 37.9 | #209 | 1 |
| [Writing & Preference](https://noometry.com/best/writing) | 35.5 | #265 | 5 |

## Strengths and weaknesses

Categories where gpt-oss-20b places highest and lowest among the models ranked in each, with its score against that category's median.

### Strongest categories

gpt-oss-20b: strongest categories
| Category | Score | vs median | Rank |
| --- | --- | --- | --- |
| [Math](https://noometry.com/best/math) | 39.4 | +2.9 | #103 of 327, top 32% |
| [Coding](https://noometry.com/best/coding) | 37.6 | −1.1 | #192 of 340, top 57% |
| [Knowledge](https://noometry.com/best/knowledge) | 34.6 | −2.7 | #195 of 314, top 63% |

### Weakest categories

gpt-oss-20b: weakest categories
| Category | Score | vs median | Rank |
| --- | --- | --- | --- |
| [Agentic & Tool Use](https://noometry.com/best/agentic) | 9.3 | −21.0 | #154 of 154, top 100% |
| [Writing & Preference](https://noometry.com/best/writing) | 35.5 | −18.2 | #265 of 312, top 85% |
| [Instruction Following](https://noometry.com/best/instruction-following) | 61.8 | −9.5 | #240 of 305, top 79% |

## Closest competitors

The models ranked just above and below gpt-oss-20b. When scores are this close, price and speed are often the better way to choose.

Models ranked closest to gpt-oss-20b
| Model | Rank | Score | Blended $/M | Speed |  |
| --- | --- | --- | --- | --- | --- |
| [Tulu 3 (Tülu 3) 70B](https://noometry.com/models/tulu-3-70b) | #251 | 33.0 | — | — | [Compare](https://noometry.com/compare/gpt-oss-20b-vs-tulu-3-70b) |
| [DeepSeek-R1-Distill-Qwen-14B](https://noometry.com/models/deepseek-r1-distill-qwen-14b) | #252 | 32.7 | — | — | [Compare](https://noometry.com/compare/deepseek-r1-distill-qwen-14b-vs-gpt-oss-20b) |
| [Qwen1.5-14B](https://noometry.com/models/qwen1-5-14b) | #253 | 32.7 | — | — | [Compare](https://noometry.com/compare/gpt-oss-20b-vs-qwen1-5-14b) |
| [Olmo 2 0325 32b Instruct](https://noometry.com/models/olmo-2-0325-32b-instruct) | #254 | 32.7 | — | — | [Compare](https://noometry.com/compare/gpt-oss-20b-vs-olmo-2-0325-32b-instruct) |
| [Laguna M.1](https://noometry.com/models/laguna-m-1) | #256 | 32.5 | — | — | [Compare](https://noometry.com/compare/gpt-oss-20b-vs-laguna-m-1) |
| [Command R+](https://noometry.com/models/command-r-plus) | #257 | 32.4 | $4.38 | — | [Compare](https://noometry.com/compare/command-r-plus-vs-gpt-oss-20b) |
| [Granite 3.1 8b Instruct](https://noometry.com/models/granite-3-1-8b-instruct) | #258 | 32.4 | — | — | [Compare](https://noometry.com/compare/gpt-oss-20b-vs-granite-3-1-8b-instruct) |
| [Pixtral Large](https://noometry.com/models/pixtral-large) | #259 | 32.2 | $3 | — | [Compare](https://noometry.com/compare/gpt-oss-20b-vs-pixtral-large) |

Sponsored placements are available on pages like this one. [Advertise on Noometry](https://noometry.com/advertise)

## Benchmark results

Every published result we track, with its source. Bold rows are the ones used for ranking; where several exist we prefer independent runs over self-reported numbers.

### Coding

gpt-oss-20b Coding benchmark results
| Benchmark | Score | Position | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [SciCode](https://noometry.com/benchmarks/scicode) | 34.4% | #101 of 121, top 84% | high | [Epoch AI](https://epoch.ai/benchmarks) |  |
| [WeirdML](https://noometry.com/benchmarks/weirdml) | 40.9% | #75 of 119, top 64% | high | [Epoch AI](https://epoch.ai/benchmarks) |  |
| [WeirdML](https://noometry.com/benchmarks/weirdml) | 36.8% |  | medium | [Epoch AI](https://epoch.ai/benchmarks) |  |
| [LMArena Coding](https://noometry.com/benchmarks/arena-coding) | 1306 | #196 of 294, top 67% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [ALE-Bench](https://noometry.com/benchmarks/ale-bench) | 566.05 | #83 of 105, top 80% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |

### Agentic & Tool Use

gpt-oss-20b Agentic & Tool Use benchmark results
| Benchmark | Score | Position | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [Terminal-Bench](https://noometry.com/benchmarks/terminal-bench) | 3.4% | #41 of 41, top 100% |  | [Epoch AI](https://epoch.ai/benchmarks) |  |

### Reasoning

gpt-oss-20b Reasoning benchmark results
| Benchmark | Score | Position | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [Kagi LLM Benchmark](https://noometry.com/benchmarks/kagi-reasoning) | 53.2% | #56 of 99, top 57% |  | [Kagi LLM Benchmark](https://help.kagi.com/kagi/ai/llm-benchmark.html) |  |
| [CritPt](https://noometry.com/benchmarks/critpt) | 1.4% | #76 of 134, top 57% | high | [Epoch AI](https://epoch.ai/benchmarks) |  |
| [Chess Puzzles](https://noometry.com/benchmarks/chess-puzzles) | 0% |  | high | [Epoch AI](https://epoch.ai/benchmarks) | 2026-08-27 |
| [Chess Puzzles](https://noometry.com/benchmarks/chess-puzzles) | 2% |  | low | [Epoch AI](https://epoch.ai/benchmarks) | 2026-08-27 |
| [Chess Puzzles](https://noometry.com/benchmarks/chess-puzzles) | 4% | #98 of 129, top 76% | medium | [Epoch AI](https://epoch.ai/benchmarks) | 2026-08-27 |
| [LMArena Hard Prompts](https://noometry.com/benchmarks/arena-hard-prompts) | 1274 | #202 of 297, top 69% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [DTBench](https://noometry.com/benchmarks/dtbench) | 68% | #97 of 151, top 65% | high | [Epoch AI](https://epoch.ai/benchmarks) |  |
| [LMCA](https://noometry.com/benchmarks/lmca) | 14.5% | #106 of 125, top 85% | high | [Epoch AI](https://epoch.ai/benchmarks) |  |
| [Epoch Capabilities Index](https://noometry.com/benchmarks/epoch-capabilities-index) | 137.82 | #116 of 213, top 55% |  | [Epoch AI](https://epoch.ai/eci) | 2025-08-05 |

### Math

gpt-oss-20b Math benchmark results
| Benchmark | Score | Position | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [OTIS Mock AIME 2024-2025](https://noometry.com/benchmarks/otis-mock-aime) | 50.8% |  | high | [Epoch AI](https://epoch.ai/benchmarks) | 2026-08-27 |
| [OTIS Mock AIME 2024-2025](https://noometry.com/benchmarks/otis-mock-aime) | 40.3% |  | low | [Epoch AI](https://epoch.ai/benchmarks) | 2026-08-27 |
| [OTIS Mock AIME 2024-2025](https://noometry.com/benchmarks/otis-mock-aime) | 65.3% | #100 of 173, top 58% | medium | [Epoch AI](https://epoch.ai/benchmarks) | 2026-08-27 |
| [Omni-MATH](https://noometry.com/benchmarks/omni-math) | 56.5% | #11 of 57, top 20% |  | [HELM Capabilities](https://crfm.stanford.edu/helm/capabilities/latest/) |  |
| [LMArena Math](https://noometry.com/benchmarks/arena-math) | 1317 | #172 of 285, top 61% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |

### Knowledge

gpt-oss-20b Knowledge benchmark results
| Benchmark | Score | Position | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [GPQA Diamond](https://noometry.com/benchmarks/gpqa-diamond) | 46% |  | high | [Epoch AI](https://epoch.ai/benchmarks) | 2026-08-06 |
| [GPQA Diamond](https://noometry.com/benchmarks/gpqa-diamond) | 53.2% |  | low | [Epoch AI](https://epoch.ai/benchmarks) | 2026-08-27 |
| [GPQA Diamond](https://noometry.com/benchmarks/gpqa-diamond) | 60.8% | #113 of 186, top 61% | medium | [Epoch AI](https://epoch.ai/benchmarks) | 2026-08-27 |
| [MMLU-Pro](https://noometry.com/benchmarks/mmlu-pro) | 74% | #28 of 58, top 49% |  | [HELM Capabilities](https://crfm.stanford.edu/helm/capabilities/latest/) |  |
| [GPQA (HELM)](https://noometry.com/benchmarks/helm-gpqa) | 59.4% | #25 of 57, top 44% |  | [HELM Capabilities](https://crfm.stanford.edu/helm/capabilities/latest/) |  |
| [LMArena Expert](https://noometry.com/benchmarks/arena-expert) | 1258 | #191 of 273, top 70% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |

### Multilingual

gpt-oss-20b Multilingual benchmark results
| Benchmark | Score | Position | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [LMArena Non-English](https://noometry.com/benchmarks/arena-non-english) | 1268 | #197 of 297, top 67% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [LMArena Chinese](https://noometry.com/benchmarks/arena-chinese) | 1314 | #183 of 285, top 65% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [LMArena German](https://noometry.com/benchmarks/arena-german) | 1255 | #173 of 231, top 75% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [LMArena Japanese](https://noometry.com/benchmarks/arena-japanese) | 1244 | #142 of 211, top 68% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [LMArena Korean](https://noometry.com/benchmarks/arena-korean) | 1236 | #149 of 213, top 70% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [LMArena Russian](https://noometry.com/benchmarks/arena-russian) | 1278 | #192 of 283, top 68% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [LMArena Spanish](https://noometry.com/benchmarks/arena-spanish) | 1267 | #175 of 226, top 78% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |

### Instruction Following

gpt-oss-20b Instruction Following benchmark results
| Benchmark | Score | Position | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [IFEval](https://noometry.com/benchmarks/ifeval) | 73.2% | #53 of 57, top 93% |  | [HELM Capabilities](https://crfm.stanford.edu/helm/capabilities/latest/) |  |
| [LMArena Instruction Following](https://noometry.com/benchmarks/arena-instruction-following) | 1236 | #224 of 298, top 76% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |

### Long Context

gpt-oss-20b Long Context benchmark results
| Benchmark | Score | Position | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [LMArena Longer Query](https://noometry.com/benchmarks/arena-longer-query) | 1250 | #222 of 291, top 77% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |

### Writing & Preference

gpt-oss-20b Writing & Preference benchmark results
| Benchmark | Score | Position | Setting | Source | Date |
| --- | --- | --- | --- | --- | --- |
| [LMArena Text](https://noometry.com/benchmarks/arena-text) | 1287 | #200 of 297, top 68% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [LMArena Creative Writing](https://noometry.com/benchmarks/arena-creative-writing) | 1201 | #232 of 295, top 79% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |
| [EQ-Bench Creative Writing](https://noometry.com/benchmarks/eqbench-creative-writing) | 666 | #112 of 115, top 98% |  | [EQ-Bench](https://eqbench.com/creative_writing.html) |  |
| [WildBench](https://noometry.com/benchmarks/wildbench) | 73.7% | #48 of 57, top 85% |  | [HELM Capabilities](https://crfm.stanford.edu/helm/capabilities/latest/) |  |
| [LMArena Multi-Turn](https://noometry.com/benchmarks/arena-multi-turn) | 1268 | #212 of 295, top 72% |  | [LMArena](https://lmarena.ai/leaderboard/text) | 2026-10-08 |

## API pricing by provider

gpt-oss-20b API prices
| Route | Input $/M | Output $/M | Cached input $/M | Checked |
| --- | --- | --- | --- | --- |
| [bedrock](https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html) | $0.07 | $0.30 | — | 2026-10-10 |
| [deepinfra](https://deepinfra.com/models) | $0.03 | $0.14 | — | 2026-10-10 |
| [groq](https://console.groq.com/docs/models) | $0.075 | $0.30 | $0.0375 | 2026-10-10 |
| [openrouter](https://openrouter.ai/openai/gpt-oss-20b) | $0.018 | $0.09 | $0.009 | 2026-10-10 |
| [vertex](https://cloud.google.com/vertex-ai/generative-ai/docs/models) | $0.07 | $0.25 | $0.007 | 2026-10-10 |

[All OpenAI API prices →](https://noometry.com/llm-pricing/openai) [Estimate your cost →](https://noometry.com/tools/cost-calculator)

## Compare gpt-oss-20b

-   [gpt-oss-20b vs Olmo 2 0325 32b Instruct](https://noometry.com/compare/gpt-oss-20b-vs-olmo-2-0325-32b-instruct)
-   [gpt-oss-20b vs Laguna M.1](https://noometry.com/compare/gpt-oss-20b-vs-laguna-m-1)
-   [gpt-oss-20b vs Qwen1.5-14B](https://noometry.com/compare/gpt-oss-20b-vs-qwen1-5-14b)
-   [gpt-oss-20b vs Command R+](https://noometry.com/compare/command-r-plus-vs-gpt-oss-20b)
-   [gpt-oss-20b vs DeepSeek-R1-Distill-Qwen-14B](https://noometry.com/compare/deepseek-r1-distill-qwen-14b-vs-gpt-oss-20b)
-   [gpt-oss-20b vs Granite 3.1 8b Instruct](https://noometry.com/compare/gpt-oss-20b-vs-granite-3-1-8b-instruct)
-   [gpt-oss-20b vs Claude Fable 5.1](https://noometry.com/compare/claude-fable-5-1-vs-gpt-oss-20b)
-   [gpt-oss-20b vs Gemini 3.8 Flash](https://noometry.com/compare/gemini-3-8-flash-vs-gpt-oss-20b)
-   [gpt-oss-20b vs Kimi K3](https://noometry.com/compare/gpt-oss-20b-vs-kimi-k3)
-   [gpt-oss-20b vs Grok 4.6](https://noometry.com/compare/gpt-oss-20b-vs-grok-4-6)
-   [gpt-oss-20b vs Qwen3.8 Max](https://noometry.com/compare/gpt-oss-20b-vs-qwen3-8-max)
-   [gpt-oss-20b vs GLM-5.3](https://noometry.com/compare/glm-5-3-vs-gpt-oss-20b)
-   [gpt-oss-20b vs Muse Spark 1.3](https://noometry.com/compare/gpt-oss-20b-vs-muse-spark-1-3)
-   [gpt-oss-20b vs DeepSeek V4 Pro](https://noometry.com/compare/deepseek-v4-pro-vs-gpt-oss-20b)

## Other OpenAI models

-   [GPT-6 Astra](https://noometry.com/models/gpt-6-astra)70.8
-   [GPT-6.1 Sol](https://noometry.com/models/gpt-6-1-sol)65.6
-   [GPT-5.6 Sol](https://noometry.com/models/gpt-5-6-sol)65.0
-   [GPT-5.5 Pro](https://noometry.com/models/gpt-5-5-pro)64.3
-   [GPT-5.5](https://noometry.com/models/gpt-5-5)63.4
-   [GPT-6 Sol](https://noometry.com/models/gpt-6-sol)61.8
-   [GPT-5.4](https://noometry.com/models/gpt-5-4)59.4
-   [GPT-5.6 Terra](https://noometry.com/models/gpt-5-6-terra)59.2

## Frequently asked questions

### How good is gpt-oss-20b?

gpt-oss-20b by OpenAI ranks 255th of 354 ranked models on the Noometry Index as of October 2026, with a score of 32.5. Its strongest category is math, where it ranks 103rd. API pricing starts at $0.018 per million input tokens and $0.09 per million output tokens, with a 131K-token context window.

### How much does gpt-oss-20b cost?

gpt-oss-20b costs $0.018 per million input tokens and $0.09 per million output tokens on openrouter, with cached input at $0.009.

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

gpt-oss-20b accepts up to 131K tokens of input and can write up to 16K tokens in one response.

### Is gpt-oss-20b open source?

Yes. gpt-oss-20b's weights are downloadable from Hugging Face (openai/gpt-oss-20b); check the license for commercial terms.

### How fast is gpt-oss-20b?

gpt-oss-20b generated about 96 output tokens per second in the Kagi LLM Benchmark's timed runs. Speed varies by provider, load and reasoning effort.

### What are gpt-oss-20b's strengths and weaknesses?

Relative to other ranked models, gpt-oss-20b places best in math, coding, knowledge and lowest in agentic & tool use, writing & preference, instruction following.

### What is gpt-oss-20b best at?

Its best category is math, where it ranks 103rd on Noometry.

### Cite this page

Noometry. (2026). gpt-oss-20b benchmarks and pricing. Retrieved October 10, 2026, from https://noometry.com/models/gpt-oss-20b

Quote Noometry with a link back to this page. It is also available in [Markdown](https://noometry.com/md/models/gpt-oss-20b.md).