gpt-oss-20b

Tool callingStructured outputReasoning

Hugging Face Launch post Website Docs

Open GPT reasoning model for self-hosted agents and controllable deployments

Context: 131K
Released: 1y ago
Parameters: 20B
Input / 1M tokens: $0.02
Output / 1M tokens: $0.09
Cached input / 1M tokens: $0.009

Sources:models.devOpenRouter

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.

Other published benchmarks

Evaluation	Reported score and source
GPQA diamond	46.0% Epoch AI · gpt-oss-20b_high Reasoning effort: high Epoch AI evaluation
GPQA diamond	53.2% Epoch AI · gpt-oss-20b_low Reasoning effort: low Epoch AI evaluation
GPQA diamond	60.8% Epoch AI · gpt-oss-20b_medium Reasoning effort: medium Epoch AI evaluation
OTIS Mock AIME 2024-2025	50.8% Epoch AI · gpt-oss-20b_high Reasoning effort: high Epoch AI evaluation
OTIS Mock AIME 2024-2025	40.3% Epoch AI · gpt-oss-20b_low Reasoning effort: low Epoch AI evaluation
OTIS Mock AIME 2024-2025	65.3% Epoch AI · gpt-oss-20b_medium Reasoning effort: medium Epoch AI evaluation
Terminal Bench	3.4% Terminal-Bench v2 Leaderboard · gpt-oss-20b Agent: Mini-SWE-Agent Terminal-Bench v2 Leaderboard Original-source terms apply
Terminal Bench	3.1% Terminal-Bench v2 Leaderboard · gpt-oss-20b Agent: Terminus 2 Terminal-Bench v2 Leaderboard Original-source terms apply
Terminal Bench	3.4% https://www.tbench.ai/leaderboard/terminal-bench/2.0 · gpt-oss-20b_unknown Agent: Mini-SWE-Agent · Reasoning effort: unknown https://www.tbench.ai/leaderboard/terminal-bench/2.0 Original-source terms apply
Terminal Bench	3.1% https://www.tbench.ai/leaderboard/terminal-bench/2.0 · gpt-oss-20b_unknown Agent: Terminus 2 · Reasoning effort: unknown https://www.tbench.ai/leaderboard/terminal-bench/2.0 Original-source terms apply

Benchmark compilation: Epoch AI, Capabilities & benchmarking (CC BY attribution). Scores are converted to percentages and model names matched to the Vertical Tabs catalog. Individual evaluations are credited to their original sources. External results may carry separate licences and evaluation configurations; do not interpret scores from different harnesses as directly comparable.

Specifications

Developer: OpenAI
Model family: GPT
Type: Language model
Predecessor: —
Successor: —
Released: August 5, 2025
Model ID: openai/gpt-oss-20b
Openness: Open weights
License: —
Parameters: 20B
Input modalities: text
Output modalities: text
Modality: text->text
Native context: 128K
Tokenizer: GPT
Max output tokens: 33K
Knowledge cutoff: 2024-06-30
Hugging Face: openai/gpt-oss-20b
Source freshness: Metadata synced Oct 8, 2026, 10:17 PM UTC

Open-weight builds

Ollama variants

Every published Ollama tag matched to this model, including parameter editions, instruction/base variants, and quantized builds.

4 builds

Variant	Parameters	Type	Quantization	Size	Context	Input
gpt-oss:20b-cloud	20B	cloud	—	—	128K	text
gpt-oss:latest	20B	—	—	14 GB	128K	text
gpt-oss:120b-cloud	120B	cloud	—	—	128K	text
gpt-oss:120b	120B	—	—	65 GB	128K	text

Additional source evidence

Source model: openai/gpt-oss-20b
Source family: gpt-oss
Last updated: 2025-08-05
Model context: 131K
Model output limit: 33K
Open weights: Yes
File attachments: No
Reasoning: Yes
Tool calling: Yes
Structured output: Yes
Temperature control: Yes

Downloadable weights

Hugging Face

Complete source record (JSON)

{
  "id": "openai/gpt-oss-20b",
  "name": "GPT OSS 20B",
  "description": "Open GPT reasoning model for self-hosted agents and controllable deployments",
  "family": "gpt-oss",
  "last_updated": "2025-08-05",
  "attachment": false,
  "reasoning": true,
  "tool_call": true,
  "structured_output": true,
  "temperature": true,
  "modalities": {
    "input": [
      "text"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 131072,
    "output": 32768
  },
  "release_date": "2025-08-05",
  "open_weights": true,
  "weights": [
    {
      "label": "Hugging Face",
      "url": "https://huggingface.co/openai/gpt-oss-20b"
    }
  ]
}

Infrastructure

Provider availability

All providers

Provider	Endpoint	Context	Input /M tokens	Output /M tokens	Cached input /M tokens	Cache write /M tokens	Uptime 1d	Data handling
Darkbloom	darkbloom/fp8	131K	$0.02	$0.09	$0.009	—	95.02%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Darkbloom	darkbloom/fp8Batch pricing	131K	$0.02	$0.09	$0.009	—	95.02%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
AkashML	akashml/fp4	131K	$0.02	$0.10	—	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
AkashML	akashml/fp4Batch pricing	131K	$0.02	$0.10	—	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
DekaLLM	dekallm/bf16	131K	$0.03	$0.14	$0.03	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
DekaLLM	dekallm/bf16Batch pricing	131K	$0.03	$0.14	$0.03	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
CoreWeave	coreweave/fp4	131K	$0.03	$0.13	$0.03	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
CoreWeave	coreweave/fp4Batch pricing	131K	$0.03	$0.13	$0.03	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
DeepInfra	deepinfra/bf16	131K	$0.03	$0.14	—	—	99.70%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
DeepInfra	deepinfra/bf16Batch pricing	131K	$0.03	$0.14	—	—	99.70%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Parasail	parasail/fp4	131K	$0.03	$0.15	$0.02	—	99.93%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Parasail	parasail/fp4Batch pricing	131K	$0.03	$0.15	$0.02	—	99.93%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
SiliconFlow	siliconflow/fp8	131K	$0.04	$0.18	—	—	89.59%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
SiliconFlow	siliconflow/fp8Batch pricing	131K	$0.04	$0.18	—	—	88.47%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Amazon Bedrock	amazon-bedrock/eu-west-1	131K	$0.07	$0.15	—	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Amazon Bedrock	amazon-bedrock	131K	$0.07	$0.15	—	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Amazon Bedrock	amazon-bedrock/eu-west-1Batch pricing	131K	$0.07	$0.15	—	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Amazon Bedrock	amazon-bedrockBatch pricing	131K	$0.07	$0.15	—	—	100.00%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Google	google-vertex/us-central1	131K	$0.07	$0.25	—	—	98.89%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Google	google-vertex/us-central1Batch pricing	131K	$0.07	$0.25	—	—	99.29%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Groq	groq	131K	$0.08	$0.30	$0.04	—	99.93%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Groq	groqBatch pricing	131K	$0.08	$0.30	$0.04	—	99.86%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15

Provider data:OpenRouter

Text provenance

EU · product-scoped

OpenAI textGrain watermarking

In the EU, ChatGPT-generated text uses textGrain. OpenAI says API customers can opt in globally for select models, so watermarking depends on the product surface and configuration.

Official explanation

Deployment graph

Can I run this?

Vertical Tabs only shows compatibility when it has structured evidence tying this exact model to hardware, runtime, quantization, memory use, and observed performance.

Browse hardware