How much VRAM does Olmo 3.1 32B need?
Olmo 3.1 32B needs 18.2 GB for its weights at 4-bit, 20.2 GB with an 8K context and 21.7 GB with 32K. At 8-bit, 33.9 GB.
- Parameters
- 32.2B
- Longest context
- 64K
- Released
- Dec 2025
- License
- Apache 2.0
- Strongest at
- Chat, Coding, Agents, Docs
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 64K |
|---|---|---|---|---|
| 16-bit | 60.0 GB | 62.1 GB | 63.6 GB | 65.6 GB |
| 8-bit | 31.9 GB | 33.9 GB | 35.4 GB | 37.4 GB |
| 6-bit | 24.6 GB | 26.7 GB | 28.2 GB | 30.2 GB |
| 5-bit | 21.3 GB | 23.4 GB | 24.9 GB | 26.9 GB |
| 4-bit | 18.2 GB | 20.2 GB | 21.7 GB | 23.7 GB |
| 3-bit | 14.7 GB | 16.7 GB | 18.2 GB | 20.2 GB |
| 2-bit | 11.3 GB | 13.3 GB | 14.8 GB | 16.8 GB |
| 1.6-bit | 6.2 GB | 8.2 GB | 9.7 GB | 11.7 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 | 4-bit | 48K | 32 tok/s | 2.0 s |
| RTX 4090 | 4-bit | 48K | 34 tok/s | 0.8 s |
| RTX 5090 | 6-bit | 64K | 44 tok/s | 0.7 s |
| 2× RTX 3090 | 8-bit | 64K | 19 tok/s | 1.1 s |
| RTX PRO 6000 Blackwell | 8-bit | 64K | 35 tok/s | 0.5 s |
| M4 Max 128 GB | 8-bit | 64K | 10.0 tok/s | 11 s |
| M5 Max 128 GB | 8-bit | 64K | 11 tok/s | 3.2 s |
| Ryzen AI Max+ 395 128 GB | 8-bit | 64K | 4.5 tok/s | 6.4 s |
| DGX Spark 128 GB | 8-bit | 64K | 5.6 tok/s | 2.2 s |
| M3 Ultra 256 GB | 8-bit | 64K | 15 tok/s | 6.2 s |
| M5 Ultra 512 GB | 8-bit | 64K | 21 tok/s | 1.9 s |
| H100 SXM | 8-bit | 64K | 61 tok/s | 0.4 s |
| DGX Station GB300 | 8-bit | 64K | 115 tok/s | 0.2 s |
| 8× H200 NVL | 8-bit | 64K | 84 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
- Ollama
ollama run olmo-3.1:32b-instruct- llama.cpp
llama-server -hf unsloth/Olmo-3.1-32B-Instruct-GGUF- vLLM
vllm serve allenai/Olmo-3.1-32B-Instruct- MLX (Mac)
mlx_lm.chat --model mlx-community/Olmo-3.1-32B-Instruct-4bit
Questions
How much VRAM does Olmo 3.1 32B need?
Olmo 3.1 32B needs 18.2 GB for its weights at 4-bit, 20.2 GB with an 8K context and 21.7 GB with 32K. At 8-bit, 33.9 GB.
Can Olmo 3.1 32B run on an RTX 4090?
Yes. At 4-bit it holds 48K of context and writes about 34 tokens a second.