USASI
Model release

gpt-oss-20b

Release in the gpt-oss family · version gpt-oss-20b

Maintained by OpenAI12

gpt-oss-20b is the smaller gpt-oss model: a 24-layer mixture-of-experts transformer with 20.9B total and about 3.6B active parameters per token, using 32 experts with the top 4 selected per token. It is text-only, supports context up to 131,072 tokens, and ships with its MoE weights quantized to MXFP4.1

Last reviewedEntry updated Documented release Aug 2025

Availability and license

Overall availability

Public

Downloadable from Hugging Face without gating.2

Availability is separate from permission: read the license before using or redistributing.

The model card says the models are available under Apache 2.0 and OpenAI's gpt-oss usage policy. The USAGE_POLICY file shipped with the weights is a short statement asking users to comply with all applicable law; it adds no further listed restrictions.14

Model-disclosure tier

Computed from the checklist below using USASI rubric v0.1. An editorial category, not a certification.
Model-disclosure tier (USASI rubric v0.1): Open-weight

The model parameters for this release can be downloaded by the public. License terms may still restrict use, redistribution, or commercial use.

How tiers are computed

Public materials checklist

Items for a model under USASI rubric v0.1. Unknown means unassessed or insufficient evidence.
Public materials checklist for gpt-oss-20b
ItemStatusNotes and evidence
WeightsCan the general public download the model parameters for this release?PublicMXFP4-quantized MoE weights plus an original-format checkpoint are in the Hugging Face repository.2
Inference codeIs code for running the model published?PublicThe GitHub repository has reference PyTorch, Triton, and Metal implementations, tools, and a Responses API server.5
Training codeIs the code used to train the model published?UnknownThe official repository covers inference, tools, and evaluation; no training code was found.
Training-data informationPublic = the training data itself can be obtained. Partial = composition or sources are documented without full access.PartialThe model card describes a text-only dataset of trillions of tokens focused on STEM, coding, and general knowledge, filtered for hazardous biosecurity content, with a June 2024 knowledge cutoff. The data itself is not released.1
Training recipeAre the training configuration and procedure documented in enough detail to follow?PartialThe model card documents architecture, tokenizer, training compute (almost 10x fewer H100-hours than the 2.1 million used for gpt-oss-120b), and post-training with chain-of-thought reinforcement learning at a high level, not in reproducible detail.1
Evaluation materialsPublic = evaluation code or prompts that let others re-run the evaluations are published. Partial = results only.PublicThe model card reports benchmark and safety results; the repository includes evaluation code adapted from simple-evals for running GPQA and HealthBench.17

What it is useful for

Presented for lower-latency, local, or specialized use cases, including tool use and function calling; the model card says it can be fine-tuned on consumer hardware.2

Run and use notes

Documented facts only. No hardware or performance claims are made without a cited source and stated assumptions.
  • The model card documents running the model with Transformers, vLLM, Ollama, and LM Studio, and OpenAI's repository provides reference PyTorch, Triton, and Metal implementations.25
  • OpenAI states that with its MoE weights in MXFP4 precision (4.25 bits per parameter), gpt-oss-20b can run on systems with as little as 16GB of memory; the model card lists the checkpoint size as 12.8 GiB.12
  • The model must be prompted with OpenAI's harmony response format; the model card says it will not work correctly otherwise.2

Organization context

Other releases in the gpt-oss family

  • gpt-oss-120bModel-disclosure tier (USASI rubric v0.1): Open-weight

gpt-oss family overview

U.S. eligibility

Project eligibility rests on documented governing or maintaining entities, not on contributors.

Eligible · basis: U.S. headquarters

Developed and published by OpenAI. Its operating company, OpenAI Group PBC, lists its address in San Francisco, California, in a February 2026 SEC-filed agreement.18

Assessed Sep 29, 2026

Sources

  1. 1.
    gpt-oss-120b & gpt-oss-20b Model Card (external site: cdn.openai.com)

    OpenAI · Model card · published Aug 5, 2025 · accessed Sep 29, 2026

  2. 2.
    openai/gpt-oss-20b (external site: huggingface.co)

    OpenAI (Hugging Face) · Model card · accessed Sep 29, 2026

  3. 3.
    openai/gpt-oss-20b LICENSE (external site: huggingface.co)

    OpenAI (Hugging Face) · License · accessed Sep 29, 2026

  4. 4.
    openai/gpt-oss-20b USAGE_POLICY (external site: huggingface.co)

    OpenAI (Hugging Face) · License · accessed Sep 29, 2026

  5. 5.
    openai/gpt-oss (GitHub repository) (external site: github.com)

    OpenAI · Repository · accessed Sep 29, 2026

  6. 6.
  7. 7.
    gpt_oss.evals README (external site: raw.githubusercontent.com)

    OpenAI · Repository · accessed Sep 29, 2026

  8. 8.
    Exhibit 10.1: Equity commitment letter agreement between OpenAI Group PBC and Amazon (external site: sec.gov)

    U.S. Securities and Exchange Commission (Amazon.com, Inc. filing) · Filing · published Feb 27, 2026 · accessed Sep 29, 2026

This listing is not an endorsement, a safety assessment, or a federal approval.

Support Us

Help keep USASI useful.

Optional. No USASI account required. Payment takes place on the linked provider’s website (Buy Me a Coffee).

About supporting this project