L4

Speed is modeled, not measured: decode is memory-bandwidth bound, so tokens per second is bytes read per token against achievable bandwidth. Mixture-of-experts models carry a wider band because only the routed experts are read each step, and few have been measured publicly.

Aggregated from community-submitted runs, so the spread is wide by nature — it covers different models, resolutions, step counts and settings, not one controlled configuration. Read the middle 50% rather than the median alone. These figures are reproduced with attribution from vladmandic-sd-data-benchmark, which publishes no licence — so we display and link rather than redistribute them.

What AI models can a L4 run?

1720 of 2118 indexed open-weight models fit a L4 at 131,072 context with q8_0 KV cache, the largest being Phi-3-medium-128k-instruct at Q4_K_M. That covers text, vision-language, image, video and speech models.

How much usable memory does a L4 actually have?

Its nameplate is 24 GB, but about 22.32 GiB is available to a model once driver and compositor overhead is accounted for.

Is a L4 fast for local AI?

Its memory bandwidth is 300 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.