Skip to content
localmodel.run

Text model · gpt-oss

gpt-oss 20B: RAM and VRAM requirements

gpt-oss 20B needs about 13.2 GB to run at Q4_K_M (the Q4_K_M GGUF file is ~11.28 GB to download; KV cache and overhead add the rest), or about 2.8 GB at Q8_0. The lightest hardware that runs it is Nvidia GeForce RTX 4060 Ti (16GB).

gpt-oss family · 21B params (Mixture-of-Experts: activates only 3.6B of 21B params per token, so generation is faster than the total size suggests) · released Aug 2025 · 10.2M Ollama pulls.

Shopping for hardware? See what runs gpt-oss 20B →

License Apache-2.0 · Commercial OK ↓ 8.5M/mo ♥ 4.9K on HuggingFace
Q4_K_M GGUF
11.28 GB
Q8_0 GGUF
0.86 GB
Memory @ Q4 (4k)
~13.2 GB
Context
128 k

Q4_K_M and Q8_0 are GGUF download (file) sizes. Memory @ Q4 includes the KV cache and overhead, so it runs larger than the file on disk.

Will it run on your device?

gpt-oss 20B runs on 17 of 40 tracked devices at Q4_K_M.

17 run well 0 tight fit 23 too small
Fit check Q4_K_M
No, not enough memory
needs 13.2 GB usable 10.5 GB
See the full breakdown

Memory breakdown

Weights (Q4_K_M)11.28 GB
+
KV cache (4k)1.1 GB
+
Overhead0.8 GB
=
Total13.2 GB

How context length changes it

4k context ~13.2 GB 32k context ~20.9 GB 128k context ~47.3 GB

Longer context grows the KV cache, which for gpt-oss 20B is sized by its 3.6B active params, not the full 21B. It needs ~13.2 GB at 4k and ~47.3 GB at 128k.

Quantization sizes

GGUF quantson disk
Quantization Size on disk
Q2_K 8.8 GB est
Q3_K_M 10.3 GB est
Q4_K_M (default) 11.28 GB
Q5_K_M 15 GB est
Q6_K 17.2 GB est
Q8_0 0.86 GB
FP16 42 GB est

Lower quant = smaller and faster, slightly lower quality. Q4_K_M is the common default.

Run it

Ollama
$ ollama run gpt-oss:20b
llama.cpp
$ llama-cli -hf ggml-org/gpt-oss-20b-GGUF:Q4_K_M
LM Studio
$ lms get ggml-org/gpt-oss-20b-GGUF

Which devices can run gpt-oss 20B?

Apple Silicon Macs

RAM-only laptops

iPhone & iPad

Android

NVIDIA GPUs

AMD GPUs

Same job, different size

Pick by what fits your memory: step down to free up VRAM, or step up if you have headroom.

Next size down DeepSeek-V2-Lite 16B Next size up Mistral Small 3 24B 24B

Similar models

Head-to-head

FAQ

How much VRAM or RAM does gpt-oss 20B need?

At Q4_K_M, gpt-oss 20B needs about 13.2 GB (weights ~11.28 GB + KV cache + overhead) at a 4k context. At Q8_0 budget ~2.8 GB.

What is the Q4_K_M GGUF file size of gpt-oss 20B?

The Q4_K_M GGUF file is about 11.28 GB to download, and the Q8_0 GGUF is about 0.86 GB. That is the weights file on disk; to run it you also need room for the KV cache and overhead, so budget ~13.2 GB of memory at Q4_K_M.

Can gpt-oss 20B run on a laptop?

gpt-oss 20B is large; you need a 24 GB+ GPU or a 32-48 GB Mac at Q4_K_M.

Is gpt-oss 20B cheaper to run because it is a MoE model?

It is faster, not lighter. gpt-oss 20B activates only 3.6B of 21B params per token (so it runs quickly), but all experts must stay in memory, so it still needs memory for the full 21B.

Can I use gpt-oss 20B commercially?

Yes. gpt-oss 20B is licensed Apache-2.0, which permits commercial use.

Understand the numbers

Short guides to the ideas behind gpt-oss 20B's memory and quant figures.

OpenAI gpt-oss 20B, MoE (3.6B active). Ships native MXFP4 4-bit; the official ggml-org GGUF is ~11.3GB (the real default, not a Q4_K_M requant). Runs on a 16GB GPU, or a Mac with 24GB+ of unified memory for working headroom.

Sources

Last validated 2026-08-03. Memory figures are estimates. See methodology.