OpenAI

Gpt Oss 20B

Gpt Oss 20B is an open-weights 20B-parameter model by OpenAI. It holds an Intelligence score of 43.3 on the local.ai leaderboard (2026-08 snapshot).

open weights
43.3
local.ai Intelligence
Parameters
20B
Benchmarks (local.ai snapshot)
τ²-bench
78%
GAIA
44%
GDPval
37%
HuggingFace metadata · updated 2025-08-26
repo
openai/gpt-oss-20b
license
apache-2.0
downloads
8,117,114
likes
4,917

Metadata refreshed from the HuggingFace API (snapshot 2026-08-15T15:53:42.405Z). Context length and long-form descriptions are the next pipeline stage.

Does it fit your hardware?

Detect or select your GPU first. Green = weights + context headroom, yellow = tight, grey = does not fit.

Q4_K_M 11.6GQ6_K 18GQ8_0 22GFP16 40G
Quantization≈ File sizeNote
IQ2_XXS5.4 GBExtreme compression, largest quality loss
IQ3_XXS7.2 GBVery small
Q3_K_M9.2 GBSmall
NVFP410 GBNVIDIA 4-bit
Q4_K_S10.8 GBCommon
Q4_K_M11.6 GBMost common sweet spot
Q5_K_M15.2 GBBalanced
Q6_K18 GBNear-lossless
Q8_022 GBNear-lossless
FP1640 GBOriginal weights

Actual files on HuggingFace

FileSize
original/model.safetensors12.82 GB
metal/model.bin12.81 GB
model-00001-of-00002.safetensors4.47 GB
model-00000-of-00002.safetensors4.46 GB
model-00002-of-00002.safetensors3.88 GB
GGUF: unsloth/gpt-oss-20b-GGUF · 564,597 downloads
FileSize
gpt-oss-20b-F16.gguf12.85 GBdownload
gpt-oss-20b-Q2_K.gguf10.68 GBdownload
gpt-oss-20b-Q2_K_L.gguf10.95 GBdownload
gpt-oss-20b-Q3_K_M.gguf10.72 GBdownload
gpt-oss-20b-Q3_K_S.gguf10.68 GBdownload
gpt-oss-20b-Q4_0.gguf10.71 GBdownload
gpt-oss-20b-Q4_1.gguf10.78 GBdownload
gpt-oss-20b-Q4_K_M.gguf10.83 GBdownload
gpt-oss-20b-Q4_K_S.gguf10.82 GBdownload
gpt-oss-20b-Q5_K_M.gguf10.91 GBdownload
gpt-oss-20b-Q5_K_S.gguf10.91 GBdownload
gpt-oss-20b-Q6_K.gguf11.21 GBdownload
GGUF: lmstudio-community/gpt-oss-20b-GGUF · 117,571 downloads
FileSize
gpt-oss-20b-MXFP4.gguf11.28 GBdownload

What you save self-hosting this model

Uses the model's own API price when it exists; otherwise the closest commercial equivalent.

Waiting for hardware detection…

RunLocal presents third-party leaderboard data (local.ai / Exo Labs, 2026-08) with its own fit and cost estimates. Verify pricing, licenses and file sizes at the source before deploying a model.