gpt-oss

gpt-oss-20b

OpenAI · open-permissive apache-2.0

Parameters (B)21.512
Architecturemoe
Layers / KV heads / head dim24 / 8 / 64
Context window131072
Modalitiesin: text; out: text
GGUFavailable
Commercial useyes

Minimum hardware (estimate)

At Q4_K_M and 8k context, total working set ≈ 14.0 GiB. On a machine with 64 GiB system RAM and no discrete GPU: CPU-only, slow. Limiting factor: No GPU VRAM; fits in system RAM but inference will be CPU-bound.

Open in fit advisor

Scores

BenchmarkValueUncertaintyRun bySource
GPQA Diamond46.0%unknownthird-partyepoch-ai
OTIS Mock AIME 2024-202553.9%unknownthird-partyepoch-ai
Terminal-Bench3.4%unknownthird-partyepoch-ai
Terminal-Bench3.4%unknownthird-partyepoch-ai
Terminal-Bench3.1%unknownthird-partyepoch-ai
Terminal-Bench3.1%unknownthird-partyepoch-ai

Provenance

HF: openai/gpt-oss-20b · snapshot 2026-08-10