vram.run Models Hardware Providers Cloud State of Inference
API provider data is live · Hardware & cloud pricing curated 2026-02-23

DeepSeek-V2.5

by deepseek-ai

236B params · text-generation · 733 likes · 85.3k downloads

DeepSeek-V2.5 is a 236B parameter model. At Q4 quantization it requires 118GB of VRAM. It fits on H200 SXM 141 GB (29 tok/s), M2 Ultra 192GB (2 tok/s) but performance may be limited.

GPU compatibility

GPUVRAMQ4 DecodeVerdict
H200 SXM 141 GB141GB29 tok/stight
M2 Ultra 192GB192GB2 tok/stight
Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0