Tesla V100 32GB for local LLMs

5 measured results from 1 builds, each linked to its source.

Can it run? DeepSeek V4.1 Flash on 2x Tesla V100 32GB Qwen3.8 27B on 2x Tesla V100 32GB Qwen3.8 Flash-Next on 2x Tesla V100 32GB

All builds with the Tesla V100 32GB

All measured results

Hide speculative-decoding results ⚡ (MTP, DFlash, DSpark and similar draft-and-verify methods give much higher tok/s on predictable text and are not comparable 1:1 with plain decode)

All Tesla V100 32GB builds: see also all GPUs · best LLMs by memory size · download the dataset.

FAQ

How fast is the Tesla V100 32GB for local LLMs?

The fastest plain-decode result we have is 42.4 tok/s on Qwen3.8 27B (Q8_0, bnk dense engine, tensor parallel, no drafting) in Dual Tesla V100 32 GB on EPYC 7402 (bnk engine). Speed depends heavily on the model, quant and engine, so compare like with like in the table. RigBase lists 5 measured results from 1 builds.

How much memory does the Tesla V100 32GB have?

32 GB per card. Memory bandwidth: 889 GB/s (measured HBM2 bandwidth reported in the bnk README).

Which models have been measured on the Tesla V100 32GB?

DeepSeek V4.1 Flash, Qwen3.8 27B, Qwen3.8 Flash-Next.

Where do these numbers come from?

Every row links to a build page with its original public source (benchmark logs, READMEs, forum posts). RigBase does not re-run the benchmarks. Results marked ⚡ use speculative decoding and are not comparable 1:1 with plain decode.