How much VRAM does Mistral Small 4 119B need?
Mistral Small 4 119B needs 67.4 GB for its weights at 4-bit, 68.4 GB with an 8K context and 68.9 GB with 32K.
- Parameters
- 119B
- Active per token
- 6.5B
- Longest context
- 256K
- Released
- Mar 2026
- License
- Apache 2.0
- Strongest at
- Chat, Docs
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| As released (FP8) | 113 GB | 114 GB | 114 GB | 116 GB |
| 6-bit | 91.2 GB | 92.2 GB | 92.7 GB | 94.8 GB |
| 5-bit | 79.1 GB | 80.1 GB | 80.6 GB | 82.7 GB |
| 4-bit | 67.4 GB | 68.4 GB | 68.9 GB | 71.1 GB |
| 3-bit | 54.3 GB | 55.3 GB | 55.9 GB | 58.0 GB |
| 2-bit | 41.7 GB | 42.7 GB | 43.2 GB | 45.3 GB |
| 1.6-bit | 22.9 GB | 23.9 GB | 24.5 GB | 26.6 GB |
One conversation, 16-bit context cache, runtime overhead included.
What runs it
| Machine | Best fit | Context | Speed | First reply |
|---|---|---|---|---|
| RTX 3060 12 GB | Too big | |||
| RTX 4060 Ti 16 GB | Too big | |||
| RTX 3090 | Too big | |||
| RTX 4090 | Too big | |||
| RTX 5090 | Too big | |||
| 2× RTX 3090 | 2-bit | 104K | 170 tok/s | 0.7 s |
| RTX PRO 6000 Blackwell | 5-bit | 256K | 171 tok/s | 0.3 s |
| M4 Max 128 GB | 5-bit | 256K | 55 tok/s | 6.5 s |
| M5 Max 128 GB | 5-bit | 256K | 60 tok/s | 1.9 s |
| Ryzen AI Max+ 395 128 GB | 5-bit | 256K | 30 tok/s | 3.8 s |
| DGX Spark 128 GB | FP8 | 48K | 27 tok/s | 2.5 s |
| M3 Ultra 256 GB | FP8 | 256K | 57 tok/s | 3.7 s |
| M5 Ultra 512 GB | FP8 | 256K | 74 tok/s | 1.1 s |
| H100 SXM | 4-bit | 256K | 260 tok/s | 0.3 s |
| DGX Station GB300 | FP8 | 256K | 287 tok/s | 0.1 s |
| 8× H200 NVL | FP8 | 256K | 243 tok/s | 0.1 s |
Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.
How to run it
- llama.cpp
llama-server -hf unsloth/Mistral-Small-4-119B-2603-GGUF- vLLM
vllm serve mistralai/Mistral-Small-4-119B-2603- MLX (Mac)
mlx_lm.chat --model mlx-community/Mistral-Small-4-119B-2603-4bit
Questions
How much VRAM does Mistral Small 4 119B need?
Mistral Small 4 119B needs 67.4 GB for its weights at 4-bit, 68.4 GB with an 8K context and 68.9 GB with 32K.
Can Mistral Small 4 119B run on an RTX 4090?
Not on one card. The smallest machine here that runs it is the 2× RTX 3090.