Requested LLM benchmark runs

Pick a model and hardware to request a measured run, or vote for one below. llm-speed measures the most-requested NVIDIA pairs first, within budget. Apple Silicon and other hardware may need a community run, so those requests count too: anyone with that machine can submit their own. Vote counts are requests, not benchmark results.

Request a run

No measured run yet for these pairs. Vote for the one you want; the most-requested runs go first. See all requests.

Codestral-22B-v0.1 on RTX 50800 votes

Get an email when this run is live (optional)Also send me new benchmark results (about monthly, unsubscribe anytime).

We never sell or share your email. Privacy

llm-speed can measure this card · Submit your own

Measurement wanted

Pairs people asked about that nobody has measured yet. Each page has the exact command to contribute a run.

DeepSeek-V4.1-Flash on 64GB system RAM · DR-005
DeepSeek-V4.1-Flash on 128GB system RAM · DR-005
Qwen3.8-Flash-Next on Apple M5 Pro 48GB · DR-002
Qwen3.8-Flash-Next on Apple M5 Max 48GB · DR-002
Qwen3.8-Flash-Next on RTX 3090 · DR-002
Qwen3.8-Flash-Next on RTX 4090 · DR-002
Qwen3.8-Flash-Next on RTX 5090 · DR-002
Qwen3.8-Flash-Next on RTX 3060 12GB · DR-002
Drex-Max v1 on RTX 4090 · DR-006
Drex-Max v1 on RTX 3090 · DR-006
Drex-Max v1 on a 16GB GPU · DR-006
Drex-Max v1 on Apple M4 · DR-006
DeepSeek-V4.1-Flash on a 12GB GPU with 64GB system RAM · OPP-0103
Qwen3.8-27B on Apple Silicon (any M-series Mac) · OPP-0106
Qwen3.8-27B on Apple M5 Max 48GB · OPP-0106
Qwen3.8-27B on Apple M5 Max 128GB · DR-018
Qwen3.8-35B-A3B on RTX 3060 12GB · OPP-0104
Qwen3.8-27B with MTP on RTX 5090 · DR-003
Qwen3.8-27B with MTP on RTX 4090 · DR-003
Qwen3.8-27B on a 24GB GPU at 64K filled context · DR-007
Qwen3.8-27B on a 24GB GPU at 128K filled context · DR-007
DeepSeek-V4.1-Flash on RTX 5090 · DR-011

Top requests

Qwen3.8-27B on Ryzen AI MAX+ 3953 votes
Needs a community run · Submit your own
Qwen2.5-32B-Instruct on RTX 40901 vote
llm-speed can measure this card · Submit your own
Qwen3-Coder-Next on RTX 40901 vote
llm-speed can measure this card · Submit your own
Qwen3.8-27B on L40S1 vote
llm-speed can measure this card · Submit your own
Qwen3.8-27B on DGX Spark1 vote
llm-speed can measure this card · Submit your own
Qwen3-Coder-30B-A3B-Instruct on M1 Max1 vote
Needs a community run · Submit your own
Qwen3.8-27B on M5 Pro1 vote
Needs a community run · Submit your own
Nemotron 3.5 Bolt 74B-A9B on A100 40GB1 vote
llm-speed can measure this card · Submit your own
Qwen3.8-27B on RTX 4070 Ti SUPER1 vote
llm-speed can measure this card · Submit your own
Gemma 4 12B IT QAT on M4 Pro1 vote
Needs a community run · Submit your own
Qwen3.8-27B on M4 Pro1 vote
Needs a community run · Submit your own