← CanIRunLocal · All hardware · All models
Yes. DeepSeek Coder V2 Lite 16B (MoE) fits in fast memory on RTX 5090 32GB at Q8_0 (18 GB needed, 31.5 GB usable), about 207 tok/s (estimate).
Hardware: RTX 5090 32GB · Model: DeepSeek Coder V2 Lite 16B (MoE)
RTX 5090 32GB has 32 GB of VRAM and 1792 GB/s of memory bandwidth. The checker holds back 0.5 GB of VRAM for the driver and assumes 16 GB of system RAM, so fast memory usable by Ollama is 31.5 GB. Partial offload may also use 60% of that system RAM. DeepSeek Coder V2 Lite 16B (MoE) (15.7B, 2.4B active, DeepSeek) at Q8_0 is a 17 GB download and needs 18 GB with headroom for about 8k tokens.
| Best quant | Q8_0 |
|---|---|
| Download size | 17 GB |
| Memory needed | 18 GB |
| Usable fast memory | 31.5 GB |
| Fit | fits in fast memory |
| Estimated speed | ~207 tok/s (estimate) |
| Command | ollama run deepseek-coder-v2:16b-lite-instruct-q8_0 |
Estimates, not benchmarks. Model sizes are ollama.com download sizes; memory needed adds headroom for ~8k tokens of context; speeds are rough figures from memory bandwidth. Data updated 2026-10-23.
A project by OnDevice Labs · llms.txt