Contributors › huppiflupp's local-LLM benchmark configs

huppiflupp's local-LLM benchmark configs

Every tracked benchmark config measured by huppiflupp: 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.

11 configs · 11 models · 2 stars · 127.1 tok/s best single-stream · huppiflupp on github.

Configs measured by huppiflupp

Model	Hardware	Quant	Backend	Decode	Context	Trust	Source
Qwen3.6-35B-A3B	Strix Halo	UD-IQ4_XS	llama.cpp/vulkan (speculative decode, mtp-2)	127.1 tok/s		Extracted, source-linked	source
Qwen3-30B-A3B-Instruct-2507	Strix Halo	IQ4_XS	llama.cpp/vulkan	94.3 tok/s (prefill 2115)		Extracted, source-linked	source
gemma-4-26B-A4B-it	Strix Halo	Q4_0	llama.cpp/vulkan	65.8 tok/s (prefill 1925)		Extracted, source-linked	source
Qwen3-VL-30B-A3B	Strix Halo	Q4_K_M	ollama	64.5 tok/s		Extracted, source-linked	source
Ornith-1.5-35B-A3B	Strix Halo	Q4_K_M	ollama	60 tok/s		Extracted, source-linked	source
Flint	Strix Halo	UD-Q4_K_XL	llama.cpp/vulkan	56.8 tok/s (prefill 1451)	65536	Extracted, source-linked	source
gpt-oss-120b	Strix Halo	MXFP4	llama.cpp/vulkan	53.7 tok/s (prefill 1151.9)		Extracted, source-linked	source
Qwen3-Coder-Next-80B-A3B	Strix Halo		ollama	51.1 tok/s		Extracted, source-linked	source
Nex N2.5 Mini 35B-A3B	Strix Halo	Q4_K_M	ollama	50.8 tok/s		Extracted, source-linked	source
Qwen3-Next 80B-A3B	Strix Halo		ollama	45.1 tok/s		Extracted, source-linked	source
Qwen3.8-Flash-Next	Strix Halo	UD-Q4_K_XL	gufo (speculative decode, mtp)	40.8 tok/s (prefill 1532)	65536	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.