unsloth
gpt oss 120b
gpt oss 120b is a 116.8B mixture-of-experts gpt-oss model from unsloth, with about 21.2B parameters active per token. Configured context length is 131,072 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters117B / 21.2B active
Context128K
Attentionhybrid
Licenseapache-2.0
36 layers · 8 KV heads · head dim 64 · 18 sliding layers · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| F16 | 60.9 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 60.0 GB | Hugging Face file bytes |
| Q8_0 | 59.0 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 58.9 GB | Hugging Face file bytes |
| Q6_K | 58.9 GB | Hugging Face file bytes |
| Q5_K_M | 58.6 GB | Hugging Face file bytes |
| Q5_K_S | 58.6 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 58.7 GB | Hugging Face file bytes |
| Q4_K_M | 58.5 GB | Hugging Face file bytes |
| Q4_K_S | 58.4 GB | Hugging Face file bytes |
| Q4_1 | 58.4 GB | Hugging Face file bytes |
| Q4_0 | 58.3 GB | Hugging Face file bytes |
| Q3_K_M | 58.3 GB | Hugging Face file bytes |
| Q3_K_S | 58.3 GB | Hugging Face file bytes |
| Q2_K | 116.8 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Won't fit | Q3_K_S | — |
| GeForce RTX 3060 12GB | Won't fit | Q3_K_S | — |
| GeForce RTX 4060 Ti 16GB | Won't fit | Q3_K_S | — |
| GeForce RTX 4090 | Won't fit | Q3_K_S | — |
| GeForce RTX 5090 | Won't fit | Q3_K_S | — |
| Apple M4 Pro | Won't fit | Q3_K_S | — |
| Apple M5 | Won't fit | Q3_K_S | — |
| CPU only · DDR5-5600 | Won't fit | Q3_K_S | — |
Smallest devices that fit it fully
- NVIDIA H100 SXM 80GB — Runs, UD-Q8_K_XL, 150–204 tok/s
- NVIDIA H100 PCIe 80GB — Runs, UD-Q8_K_XL, 90–122 tok/s
- NVIDIA A100 80GB SXM — Runs, UD-Q8_K_XL, 91–124 tok/s
- NVIDIA A100 80GB PCIe — Runs, UD-Q8_K_XL, 87–118 tok/s
- Apple M3 Ultra — Tight fit, UD-Q8_K_XL, 37–50 tok/s
- NVIDIA H200 SXM — Runs great, UD-Q8_K_XL, 215–293 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (Q3_K_S).
llama.cpp
llama-server -hf unsloth/gpt-oss-120b-GGUF:Q3_K_S -c 2048 -ngl 0
No matching name in the Ollama library snapshot.
LM Studio
lms get unsloth/gpt-oss-120b-GGUF # Then load quant Q3_K_S