gpt-oss-20b

Smaller gpt-oss model for local and more accessible open-weight deployments.

Editorial review

Reviewed byOpenSourcesAI EditorialLast updatedAugust 2026SourcesOpenSourcesAI checker catalog + Hugging Face model card (openai/gpt-oss-20b)

Model checkpoints, context windows, provider support, local runtime compatibility, and license terms can change quickly. Verify the exact model card before production or commercial use.

Best for

Builders evaluating more accessible open-weight deployments for local apps, prototypes, and controlled workflows.

Who should use it

Builders evaluating more accessible open-weight deployments for local apps, prototypes, and controlled workflows.
Builders who want local or self-hosted testing options.

Common workflows

Accessible local and self-hosted open-weight deployments
open weights workflows
local workflows
self-hosted workflows
developer-controlled workflows

Deployment and hardware notes

Likely more practical for local evaluation than larger gpt-oss releases, depending on quantization and runtime support.

License and usage notes

Apache 2.0. Open weights. Verify the exact model card and license terms for the checkpoint or hosted provider you use.

Strengths

Open weights model option for gpt-oss workflows.
Builders evaluating more accessible open-weight deployments for local apps, prototypes, and controlled workflows.
A smaller gpt-oss entry is worth evaluating for local workflows when compatible runtimes and quantized builds are available.
Tracked as Practical local in the OpenSourcesAI model directory.

Limitations

Smaller models are easier to run but still need task-specific evaluation, license review, and runtime checks.
Likely more practical for local evaluation than larger gpt-oss releases, depending on quantization and runtime support.
Context window and limits: 128,000 tokens.
Verify the exact model card, provider docs, license, and serving support before production use.

Local workflow notes

A smaller gpt-oss entry is worth evaluating for local workflows when compatible runtimes and quantized builds are available.

Local runtimes: Ollama where supported, LM Studio where supported, Transformers

Platforms: Windows, macOS, Linux

Will GPT-OSS 20B run on your machine?

GPT-OSS 20B is 20B parameters and needs 14 GB of VRAM at Q4_K_M — 12.5 GB of weights plus 1.5 GB of runtime overhead for the inference server itself.

VRAM by quantization

Quantization	Weights	Needs (with overhead)	Quality
Q4_K_M	12.5 GB	14 GB	good

Fit on common hardware at Q4_K_M

Hardware	Memory the model can use	System RAM	Verdict
CPU Only	None (CPU only)	16 GB	Too large
RTX 4060 Laptop	8 GB	16 GB	Too large
RTX 3060 (12GB)	12 GB	32 GB	CPU offload
RTX 4060 Ti (16GB)	16 GB	32 GB	Comfortable
RTX 3090	24 GB	64 GB	Comfortable
Apple Silicon (Unified Memory) 36 GB	27 GB of 36 GB	36 GB	Comfortable

Comfortable means VRAM clears the requirement by 2 GB or more. Tight means it covers the requirement with no margin. CPU offload means the model does not fit in VRAM but system RAM is at least 1.6× the weights, so it will run at reduced speed — expect roughly 1–5 tokens per second. Figures are weights plus a fixed runtime overhead and exclude KV-cache growth, which scales with context length.

Apple Silicon shares one pool of memory between the system and the GPU, so a model cannot use all of it. These rows apply the same 75% usable fraction the Compatibility Checker uses, which is why a 36 GB Mac is graded on less than 36 GB.

Check GPT-OSS 20B against your exact hardware How these verdicts are calculated