Contributors › xthomaswang's local-LLM benchmark configs
xthomaswang's local-LLM benchmark configs
Every tracked benchmark config measured by xthomaswang: model, quant, backend, tokens per second, source-linked and trust-tiered.
Snapshot 2026-10-29 · 413 configs on AI boxes and 288 on GPU cards · 121 models. Every number is generated from the tracker snapshot and linked to where it was measured; nothing here is typed by hand.
3 configs · 1 models · 41.5 tok/s best single-stream · xthomaswang on github.
Configs measured by xthomaswang
| Model | Hardware | Quant | Backend | Decode | Context | Trust | Source |
|---|---|---|---|---|---|---|---|
| Qwen3.8-27B | Mac Max | llama.cpp (speculative decode, mtp) | 41.5 tok/s | 16000 | Extracted, source-linked | source | |
| Qwen3.8-27B | Mac Max | mlx | 27.4 tok/s (prefill 765) | 16000 | Extracted, source-linked | source | |
| Qwen3.8-27B | Mac Max | vllm/metal | 26.2 tok/s (prefill 614) | 16000 | Extracted, source-linked | source |
More: all configs · hardware · news · contributors · methodology · llms.txt · llms-full.txt · API (OpenAPI) · JSON snapshot. Built by Altronis, private on-prem AI, Singapore.