gpt-oss-120b by OpenAI: Open-Weight Model

Runs On

What it takes to serve gpt-oss-120b (116.8B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

Precision	Weights	Memory needed	Cheapest setup	Per hour	Also fits
16-bit	233.7 GB	280.4 GB	1x MI355X (288 GB) Vultr	$2.59	2x MI300X $3.70 · 2x MI325X $4.00
8-bit	116.8 GB	140.2 GB	1x MI300X (192 GB) Vultr	$1.85	1x MI325X $2.00 · 1x MI355X $2.59
4-bit	58.4 GB	70.1 GB	1x MI300X (192 GB) Vultr	$1.85	1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.

SAVRN's Notes on gpt-oss-120b

Memory decides the hardware. gpt-oss-120b carries 116.8 billion parameters; at 16-bit the working footprint is 280.4 GB, meaning one MI355X with 288 GB at $2.59 an hour. At 8-bit it drops to 140.2 GB and one MI300X with 192 GB at $1.85 an hour covers it, and 4-bit needs 70.1 GB on the same card. For general text generation with a 131,072-token window, we would start on the 8-bit single card.

Apache 2.0 permits commercial use, modification and redistribution, provided the license and copyright notices stay attached and significant changes are stated, plus an express patent grant. Before committing, read the model card at arXiv:2508.10925 and price the hosted route: on the Index, DeepInfra lists $0.037 in and $0.17 out per million tokens, Cerebras $0.35 and $0.75. Once a month of tokens at those rates costs more than an MI300X at $1.85 an hour, run it yourself.

SAVRN Research, 2026-09-18

Model Card

By OpenAI, published under apache-2.0, revision b5c939de8f75.

Try gpt-oss · Guides · Model card · OpenAI blog

Welcome to the gpt-oss series, OpenAI’s open-weight models designed for powerful reasoning, agentic tasks, and versatile developer use cases.

We’re releasing two flavors of these open models: - gpt-oss-120b — for production, general purpose, high reasoning use cases that fit into a single 80GB GPU (like NVIDIA H100 or AMD MI300X) (117B parameters with 5.1B active parameters) - gpt-oss-20b — for lower latency, and local or specialized use cases (21B parameters with 3.6B active parameters)

Both models were trained on our harmony response format and should only be used with the harmony format as it will not work correctly otherwise.

[!NOTE] This model card is dedicated to the larger gpt-oss-120b model. Check out gpt-oss-20b for the smaller model.

Highlights

Read the full model card (735 words)

Configuration

Architecture: GptOssForCausalLM
Context length (tokens): 131,072
Layers: 36
Hidden size: 2,880
Feed-forward size: 2,880
Attention heads: 64
Key/value heads: 8
Head dimension: 64
Vocabulary size: 201,088
Experts: 128
Experts active per token: 4
Sliding window (tokens): 128
RoPE base: 150,000
Model type: gpt_oss
Quantization: mxfp4

Identity and Version

Repository: openai/gpt-oss-120b
Publisher: OpenAI
Task: Text generation
Modality: Text
Library: transformers
Parameters: 116.8B parameters
Languages: Not stated by the source
Revision: b5c939de8f754692c1647ca79fbf85e8c1e70f8a
First published: 2025-08-04
Last updated: 2025-08-26

Files and Weights

37 files, 195.8 GB in total. The weights are 23 files totalling 195.7 GB in bin, safetensors.

Weights23 files · 195.7 GB

Configuration7 files · 114.7 KB

Tokenizer2 files · 27.9 MB

Documentation2 files · 18.5 KB

Other2 files · 16.9 KB

Repository1 file · 1.6 KB

Every file

File	Type	Size	SHA-256
metal/model.bin	Weights	65.2 GB	0f3d5b8a213f
model-00000-of-00014.safetensors	Weights	4.6 GB	695218884684
model-00001-of-00014.safetensors	Weights	4.1 GB	a881aa5f561b
model-00002-of-00014.safetensors	Weights	4.6 GB	022478dd0439
model-00003-of-00014.safetensors	Weights	4.1 GB	47aee9e7b9d5
model-00004-of-00014.safetensors	Weights	4.6 GB	f6c2752acda6
model-00005-of-00014.safetensors	Weights	4.1 GB	0c8dd401544c
model-00006-of-00014.safetensors	Weights	4.6 GB	28d839f2e027
model-00007-of-00014.safetensors	Weights	4.1 GB	c8958c5f183c
model-00008-of-00014.safetensors	Weights	4.6 GB	bf1f2a88868f
model-00009-of-00014.safetensors	Weights	4.2 GB	f72d34a40042
model-00010-of-00014.safetensors	Weights	4.6 GB	f48c867c2cb0
model-00011-of-00014.safetensors	Weights	4.1 GB	a06851b2cfd3
model-00012-of-00014.safetensors	Weights	4.1 GB	3af33667c307
model-00013-of-00014.safetensors	Weights	4.6 GB	bcbcb74b043e
model-00014-of-00014.safetensors	Weights	4.1 GB	54b1be160969
original/model--00001-of-00007.safetensors	Weights	10.5 GB	68a8dc1f8e2e
original/model--00002-of-00007.safetensors	Weights	10.5 GB	19b8f0d5c7dc
original/model--00003-of-00007.safetensors	Weights	10.5 GB	0dbccd746d50
original/model--00004-of-00007.safetensors	Weights	10.5 GB	bcc73cf6d18f
original/model--00005-of-00007.safetensors	Weights	10.5 GB	15fd69843e9c
original/model--00006-of-00007.safetensors	Weights	10.4 GB	3aedef2ee0a5
original/model--00007-of-00007.safetensors	Weights	2.3 GB	20d5dfcad1ed
config.json	Configuration	2.1 KB	—
generation_config.json	Configuration	177 B	—
model.safetensors.index.json	Configuration	54.5 KB	—
original/config.json	Configuration	377 B	—
original/dtypes.json	Configuration	19.7 KB	—
original/model.safetensors.index.json	Configuration	37.8 KB	—
special_tokens_map.json	Configuration	98 B	—
LICENSE	Documentation	11.4 KB	—
README.md	Documentation	7.1 KB	—
USAGE_POLICY	Other	201 B	—
chat_template.jinja	Other	16.7 KB	—
.gitattributes	Repository	1.6 KB	—
tokenizer.json	Tokenizer	27.9 MB	0614fe83cada
tokenizer_config.json	Tokenizer	4.2 KB	—

License and Download

License: apache-2.0
Access: Open weights, no gate
Download size: 195.7 GB

Download from OpenAI

Released by OpenAI through its official repository on Hugging Face. Read the license.

Built From

Described by arXiv:2508.10925

Evaluations

Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical.

Benchmark	Conditions	Result	Reported by	Revision	Date
Idavidrein/gpqa	Task diamondMetric diamondSetup GPQA DiamondComparison conditions not established	80.8081	EvalEval Reported by a third party	Evaluated revision not stated	2026-04-16
Idavidrein/gpqa	Task diamondMetric diamondSetup Reasoning: mediumComparison conditions not established	73.1	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
Idavidrein/gpqa	Task diamondMetric diamondSetup Reasoning: high, With toolsComparison conditions not established	80.9	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
Idavidrein/gpqa	Task diamondMetric diamondSetup Reasoning: highComparison conditions not established	80.1	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
Idavidrein/gpqa	Task diamondMetric diamondSetup Reasoning: low, With toolsComparison conditions not established	68.1	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
Idavidrein/gpqa	Task diamondMetric diamondSetup Reasoning: medium, With toolsComparison conditions not established	73.5	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
Idavidrein/gpqa	Task diamondMetric diamondSetup Reasoning: lowComparison conditions not established	67.1	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
LEXam-Benchmark/LEXam	Task mcq_4_choicesMetric mcq_4_choicesComparison conditions not established	47.71	LEXam Leaderboard Reported by a third party	Evaluated revision not stated	2026-06-02
LEXam-Benchmark/LEXam	Task open_questionMetric open_questionComparison conditions not established	51.74	LEXam Leaderboard Reported by a third party	Evaluated revision not stated	2026-06-02
SWE-bench/SWE-bench_Verified	Task swe_bench_%_resolvedMetric swe_bench_%_resolvedSetup Reasoning: lowComparison conditions not established	47.9	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
SWE-bench/SWE-bench_Verified	Task swe_bench_%_resolvedMetric swe_bench_%_resolvedSetup Reasoning: mediumComparison conditions not established	52.6	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
SWE-bench/SWE-bench_Verified	Task swe_bench_%_resolvedMetric swe_bench_%_resolvedSetup Reasoning: highComparison conditions not established	62.4	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
ScaleAI/SWE-bench_Pro	Task SWE_Bench_ProMetric SWE_Bench_ProComparison conditions not established	16.2	SWE-Bench Pro official evaluation results Reported by a third party	Evaluated revision not stated	2026-02-28
TIGER-Lab/MMLU-Pro	Task mmlu_proMetric mmlu_proComparison conditions not established	80.8	EvalEval Reported by a third party	Evaluated revision not stated	2026-06-30
cais/hle	Task hleMetric hleSetup Reasoning: lowComparison conditions not established	5.2	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
cais/hle	Task hleMetric hleSetup Reasoning: medium, With toolsComparison conditions not established	11.3	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
cais/hle	Task hleMetric hleSetup Reasoning: mediumComparison conditions not established	8.6	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
cais/hle	Task hleMetric hleSetup Reasoning: highComparison conditions not established	14.9	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
cais/hle	Task hleMetric hleSetup Reasoning: high, With toolsComparison conditions not established	19	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
cais/hle	Task hleMetric hleSetup Reasoning: low, With toolsComparison conditions not established	9.1	GPT-OSS Model Card Reported by a third party	Evaluated revision not stated	2025-08-05
joelniklaus/LEXam-hard	Task lexam_hardMetric lexam_hardSetup lighteval, LEXam paper prompts, one response per question, no tools; DeepSeek-R1-0528 judge; mean of the German and English means over the 518 questions, 0-100Comparison conditions not established	37.55	SwissLegalEvals per-sample details (lighteval) Reported by a third party	Evaluated revision not stated	2026-06-11

Memory Requirements

Precision	Weights in memory
As published	195.7 GB
16-bit	233.7 GB
8-bit	116.8 GB
4-bit	58.4 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Hosted Prices

Host	Input / output	Unit	Observed
Baseten	$0.10 / $0.50	input / output, per million tokens	Sep 18, 2026
Cerebras	$0.35 / $0.75	input / output, per million tokens	Sep 18, 2026
Crusoe	$0.05 / $0.20	input / output, per million tokens	Sep 10, 2026
DeepInfra	$0.037 / $0.17	input / output, per million tokens	Sep 18, 2026
Fireworks	$0.15 / $0.60	input / output, per million tokens	Sep 18, 2026
Fireworks	$0.15 / $0.60	input / output, per million tokens	Sep 18, 2026
Groq	$0.15 / $0.75	input / output, per million tokens	Sep 18, 2026
Nebius Token Factory	$0.15 / $0.60	input / output, per million tokens	Sep 10, 2026
Novita	$0.05 / $0.25	input / output, per million tokens	Sep 18, 2026
Nscale	$0.10 / $0.40	input / output, per million tokens	Sep 18, 2026
OVHcloud	$0.09 / $0.47	input / output, per million tokens	Sep 18, 2026
SambaNova	$0.22 / $0.59	input / output, per million tokens	Sep 10, 2026
Scaleway	$0.17 / $0.68	input / output, per million tokens	Sep 18, 2026
Together AI	$0.15 / $0.60	input / output, per million tokens	Sep 18, 2026

From the SAVRN Index.

Compare gpt-oss-120b

Questions About gpt-oss-120b

How much GPU memory does gpt-oss-120b need?

About 280.4 GB at 16-bit and 70.1 GB at 4-bit: the weights (116.8B parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run gpt-oss-120b on?

At 16-bit, 1x MI355X from $2.59 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use gpt-oss-120b commercially?

Yes. gpt-oss-120b is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

What is gpt-oss-120b's context length?

131,072 tokens, from the maximum position embeddings in its published configuration.

Similar Models

Model · Text generation

NVIDIA-Nemotron-3-Super-120B-A12B-BF16

NVIDIA

For more details on how to deploy and use the model - see the Quick Start Guide below! The post-training data has a cutoff date of February 2026. The pre-training data has a cutoff date of June 2025. NVIDIA Nemotron™ is a family of open models with open weights, training data, and recipes, delivering leading efficiency and accuracy for building specialized AI agents. Nemotron-3-Super-120B-A12B-BF16 is a large language model (LLM) trained by NVIDIA, designed to deliver strong agentic, reasoning, and conversational capabilities. It is optimized for collaborative agents and high-volume workloads such as IT ticket automation. Like other models in the family, it responds to user queries and…

Open weights other 123.6B parameters 262,144 tokens transformers

View model

Model · Text generation

Qwen3-Coder-Next-FP8

Qwen

Today, we're announcing Qwen3-Coder-Next-FP8, an open-weight language model designed specifically for coding agents and local development. It features the following key enhancements: Qwen3-Coder-Next-FP8 has the following features: NOTE: This model supports only non-thinking mode and does not generate blocks in its output. Meanwhile, specifying enablethinking=False is no longer required. For more details, including benchmark evaluation, hardware requirements, and inference performance, please refer to our blog, GitHub, and Documentation. We advise you to use the latest version of transformers. The following contains a code snippet illustrating how to use the model generate content based on…

Open weights apache-2.0 79.7B parameters 262,144 tokens transformers

View model

Model · Text generation

Qwen-72B

Qwen

通义千问-72B（Qwen-72B）是阿里云研发的通义千问大模型系列的720亿参数规模的模型。Qwen-72B是基于Transformer的大语言模型, 在超大规模的预训练数据上进行训练得到。预训练数据类型多样，覆盖广泛，包括大量网络文本、专业书籍、代码等。同时，在Qwen-72B的基础上，我们使用对齐机制打造了基于大语言模型的AI助手Qwen-72B-Chat。本仓库为Qwen-72B的仓库。通义千问-72B（Qwen-72B）主要有以下特点： 1. 大规模高质量训练语料：使用超过3万亿tokens的数据进行预训练，包含高质量中、英、多语言、代码、数学等数据，涵盖通用及专业领域的训练语料。通过大量对比实验对预训练语料分布进行了优化。 2. 强大的性能：Qwen-72B在多个中英文下游评测任务上（涵盖常识推理、代码、数学、翻译等），效果显著超越现有的开源模型。具体评测结果请详见下文。 3. 覆盖更全面的词表：相比目前以中英词表为主的开源模型，Qwen-72B使用了约15万大小的词表。该词表对多语言更加友好，方便用户在不扩展词表的情况下对部分语种进行能力增强和扩展。 4. 较长的上下文支持：Qwen-72B支持32k的上下文长度。 Qwen-72B is the 72B-parameter version of the large language model series, Qwen (abbr. Tongyi Qianwen), proposed by Alibaba Cloud. Qwen-72B is a Transformer-based large…

Open weights other 72.3B parameters 32,768 tokens transformers

View model

Model · Text generation

Llama-3.3-70B-Instruct

Meta Llama

The Meta Llama 3.3 multilingual large language model (LLM) is an instruction tuned generative model in 70B (text in/text out). The Llama 3.3 instruction tuned text only model is optimized for multilingual dialogue use cases and outperforms many of the available open source and closed chat models on common industry benchmarks. Model Architecture: Llama 3.3 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety. Supported languages: English, German, French, Italian, Portuguese, Hindi, Spanish, and…

Access requested at publisher llama3.3 70.6B parameters transformers

View model

Model · Text generation

DeepSeek-V4-Flash-DSpark

DeepSeek

Note: DeepSeek-V4-Flash-DSpark is not a new model. It is the same checkpoint with an additional speculative decoding module attached. A minimal inference example is available in the inference folder. For more details, refer to: https://github.com/deepseek-ai/DeepSpec We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models — DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) — both supporting a context length of one million tokens. DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: 1. Hybrid Attention Architecture: We design a hybrid…

Open weights mit 165.3B parameters 1,048,576 tokens transformers

View model

Model · Text generation

NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4

NVIDIA

For more details on how to deploy and use the model - see the Quick Start Guide below! The post-training data has a cutoff date of February 2026. The pre-training data has a cutoff date of June 2025. NVIDIA Nemotron™ is a family of open models with open weights, training data, and recipes, delivering leading efficiency and accuracy for building specialized AI agents. Nemotron-3-Super-120B-A12B-NVFP4 is a large language model (LLM) trained by NVIDIA, designed to deliver strong agentic, reasoning, and conversational capabilities. It is optimized for collaborative agents and high-volume workloads such as IT ticket automation. Like other models in the family, it responds to user queries and…

Open weights other 67.2B parameters 262,144 tokens transformers

View model