Hardware FAQ
Hardware-specific FAQs
One page per top accelerator. Each answers the questions local-LLM enthusiasts ask: fastest model, supported size classes, backend rankings, quantization guidance — all anchored to signed runs.
RTX 5090 FAQ
165 signed runs on the leaderboard. Q&A covers the fastest measured LLM, supported 7B / 30B / 70B classes, backend comparisons, and quant guidance — all linked to
/r/<id>permalinks.M3 Ultra FAQ
26 signed runs on the leaderboard. Q&A covers the fastest measured LLM, supported 7B / 30B / 70B classes, backend comparisons, and quant guidance — all linked to
/r/<id>permalinks.RTX 4090 FAQ
14 signed runs on the leaderboard. Q&A covers the fastest measured LLM, supported 7B / 30B / 70B classes, backend comparisons, and quant guidance — all linked to
/r/<id>permalinks.M3 Pro FAQ
11 signed runs on the leaderboard. Q&A covers the fastest measured LLM, supported 7B / 30B / 70B classes, backend comparisons, and quant guidance — all linked to
/r/<id>permalinks.M4 Max FAQ
5 signed runs on the leaderboard. Q&A covers the fastest measured LLM, supported 7B / 30B / 70B classes, backend comparisons, and quant guidance — all linked to
/r/<id>permalinks.
See also the project FAQ for methodology, backend comparisons, and citation guidance.