~/model/huggingfaceopenai/gpt-oss-20b
OpenAI

gpt-oss 20B

OpenAI's August 2025 open-weight release. MoE; ships MXFP4-quantized at ~13GB on consumer cards.

Parameters
20.9B
Family
OpenAI
Context
128K tokens
FP16 weights
42GB
// where you can run it
gpt-oss:20bLM StudiovLLMMLXoMLX
// hugging face stats (cached daily)
7.9M downloads · 5K likes · license: apache-2.0 · updated 8 months ago

What you need to run this.

$ ./vrambudget --model gpt-oss-20b --by quant
FP1642 GB
M2 Max 6464 GB · fitsM3 Max 6464 GB · fitsM4 Pro 6464 GB · fitsM5 Pro 6464 GB · fits
Q8_022 GB
RTX 509032 GB · fitsRTX A600048 GB · fitsRTX 6000 Ada48 GB · fitsL40S48 GB · fits
Q6_K17 GB
RTX 309024 GB · fitsRTX 3090 Ti24 GB · fitsRTX 409024 GB · fitsRX 7900 XTX24 GB · fits
Q5_K_M14 GB
RTX 309024 GB · fitsRTX 3090 Ti24 GB · fitsRTX 409024 GB · fitsRX 7900 XTX24 GB · fits
Q4_K_M12 GB
RTX 4060 Ti 16GB16 GB · fitsRTX 4070 Ti Super16 GB · fitsRTX 408016 GB · fitsRTX 4080 Super16 GB · fits

// budgets shown at ctx 8K, concurrency 1, 15% safety headroom. Tune in the calculator →

Alternatives at this size.

$ grep --params similar catalog.json
$ similarMistral Small 324BMistral$ similarGemma 4 26B A4B26BGoogle$ similarStarCoder2 15B15BBigCode$ similarQwen 3.6 27B27BAlibaba

Discussion.

$ gh discussion list

// sign in with github to leave a comment. threads live in the repo's discussions tab.