selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

tencent logo

Can the Tesla P40 run WeMM-Video-9B?

WeMM-Video-9B·10.5B·Tesla P40·24 GB VRAM

Saved in this browser, used on every model page.
Yes

Yes. The Tesla P40 runs WeMM-Video-9B fully in VRAM at FP16 / BF16 (about 22 GB of 24 GB), at roughly 9.1 tokens/s.

best quant
FP16 / BF16
needs
22 GB
usable
24 GB
est. speed
~9.1 tok/s

Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the WeMM-Video-9B page →

Every quantisation of WeMM-Video-9B on the Tesla P40

QuantQualityMemoryFitSpeed
Q2_K Heavy loss4.9 GBfits~55 tok/s
Q3_K_M Noticeable loss6.5 GBfits~37 tok/s
Q4_K_M Small loss7.7 GBfits~30 tok/s
Q5_K_M Near-lossless8.8 GBfits~26 tok/s
Q6_K Near-lossless9.9 GBfits~22 tok/s
Q8_0 Lossless12 GBfits~17 tok/s
FP16 / BF16 pickLossless22 GBfits~9.1 tok/s

Speed is a bandwidth-based estimate for generating text (347 GB/s); prompt processing and long contexts are slower.

Measured speed on the Tesla P40

No community reports for this pair yet.

Report your own speed →

Vision models that fit the Tesla P40 comfortably

All that fit →
ModelParamsBest quantMemorySpeedLicence
9.7BQ8_0≈11 GBFits in memory~19 tok/sCommercial OK
8.8BQ8_0≈10 GBFits in memory~20 tok/sCommercial OK
12.2BQ8_0≈16 GBFits in memory~15 tok/sConditions
8.3BQ8_0≈8.9 GBFits in memory~22 tok/sCommercial OK
9.4BQ8_0≈11 GBFits in memory~19 tok/sCommercial OK
10.7BQ8_0≈12 GBFits in memory~17 tok/sConditions

Other GPUs that run WeMM-Video-9B

Frequently asked questions

Can the Tesla P40 run WeMM-Video-9B?
Yes. The Tesla P40 runs WeMM-Video-9B fully in VRAM at FP16 / BF16 (about 22 GB of 24 GB), at roughly 9.1 tokens/s.
How fast does WeMM-Video-9B run on the Tesla P40?
An estimate from memory bandwidth gives about 9.1 tokens/s at FP16 / BF16.
Which quantisation of WeMM-Video-9B should I use on the Tesla P40?
FP16 / BF16: the largest one that fits. Expected quality: lossless compared with the full-precision weights.

Memory and speed are estimates from the model's size and the card's memory bandwidth; real numbers depend on the runtime, drivers and context length. Community figures are medians of user reports.