User profile

flash-attn-fan

17 runsFirst seen 2026-04-10Avg 67.0

Best PipelineScore

88.1MAINLINE

on DeepSeek V3 671B-A37B

Total tokens93.0KAcross every task this user has run

Avg latency960msPer task, across all submissions

Tasks run42517 submissions x ~25 tasks

Rigs used14Distinct hardware tags

Category signature

Average score per category across all 17 runs.

Code

66.2

Reason

67.6

Write

66.1

Tool Use

67.9

RAG

67.8

Speed

67.0

Hardware mix

Rigs this user benchmarked on.

rtx-4080-16gb2 (12%)

ryzen-7950x-cpu-only2 (12%)

m1-air-8gb2 (12%)

b200-192gb1 (6%)

m3-max-128gb1 (6%)

a100-80gb1 (6%)

a100-40gb1 (6%)

h100-80gb1 (6%)

rtx-4090-24gb1 (6%)

ryzen-7950x-rtx-30901 (6%)

m2-pro-16gb1 (6%)

m2-ultra-192gb1 (6%)

m3-pro-36gb1 (6%)

m2-air-16gb1 (6%)

Provider mix

Where they spend their tokens.

mistral5 (29%)

alibaba2 (12%)

nous2 (12%)

deepseek1 (6%)

google1 (6%)

zhipu1 (6%)

tii1 (6%)

openbmb1 (6%)

microsoft1 (6%)

bigcode1 (6%)

meta1 (6%)

Models tried

Best score per model. Click a model to see its full page.

#	Model	Provider	Best Score	Tier	Achieved
1	DeepSeek V3 671B-A37B	deepseek	88.1	MAINLINE	2026-04-30
2	Qwen 3 32B Instruct	alibaba	82.3	MAINLINE	2026-05-10
3	Hermes 3 Llama 3.1 405B	nous	81.5	MAINLINE	2026-05-06
4	Devstral Small 24B	mistral	79.3	MAINLINE	2026-05-04
5	Mixtral 8x22B Instruct	mistral	78.3	MAINLINE	2026-05-11
6	Gemma 3 27B IT	google	77.7	MAINLINE	2026-04-24
7	Mistral Small 24B Instruct	mistral	70.8	FEEDER	2026-05-14
8	GLM 4 9B Chat	zhipu	69.4	FEEDER	2026-04-12
9	Falcon 3 10B Instruct	tii	69.2	FEEDER	2026-04-10
10	Mixtral 8x7B Instruct	mistral	68.4	FEEDER	2026-05-07
11	MiniCPM-V 2.6 8B	openbmb	66.0	FEEDER	2026-04-21
12	Hermes 3 Llama 3.1 8B	nous	64.7	FEEDER	2026-05-14
13	Mistral 7B Instruct v0.3	mistral	62.7	FEEDER	2026-05-23
14	Phi 3.5 Mini	microsoft	59.4	TAP	2026-04-27
15	StarCoder2 3B	bigcode	47.6	TAP	2026-04-11
16	Llama 3.2 1B Instruct	meta	41.9	TAP	2026-05-07
17	Qwen 2.5 0.5B Instruct	alibaba	32.2	DRIP	2026-04-30

All submissions

Every run, ordered by score.

#	Model	Hardware	Score	Tier	Code	Reason	Write	Tool Use	RAG	Speed	Tokens	Avg ms	Date
1	DeepSeek V3 671B-A37B	b200-192gb	88.1	MAINLINE	82.5	89.6	93.3	89.1	90.7	86.8	6.0K	853	2026-04-30
2	Qwen 3 32B Instruct	m3-max-128gb	82.3	MAINLINE	80.8	81.3	86.9	89.4	79.4	75.6	5.8K	1529	2026-05-10
3	Hermes 3 Llama 3.1 405B	a100-80gb	81.5	MAINLINE	84.8	82.0	75.1	83.0	76.0	85.4	5.2K	892	2026-05-06
4	Devstral Small 24B	a100-40gb	79.3	MAINLINE	76.9	79.9	79.1	83.4	76.0	81.7	5.2K	605	2026-05-04
5	Mixtral 8x22B Instruct	h100-80gb	78.3	MAINLINE	74.3	82.7	73.1	83.5	80.8	77.1	5.5K	923	2026-05-11
6	Gemma 3 27B IT	rtx-4090-24gb	77.7	MAINLINE	82.6	79.2	75.0	75.7	78.6	70.8	5.8K	1466	2026-04-24
7	Mistral Small 24B Instruct	ryzen-7950x-rtx-3090	70.8	FEEDER	68.1	73.4	69.4	74.2	69.2	71.4	5.6K	1466	2026-05-14
8	GLM 4 9B Chat	m2-pro-16gb	69.4	FEEDER	72.5	65.1	67.1	65.5	72.4	74.5	6.0K	773	2026-04-12
9	Falcon 3 10B Instruct	rtx-4080-16gb	69.2	FEEDER	68.1	70.0	64.9	72.7	65.0	74.5	4.9K	767	2026-04-10
10	Mixtral 8x7B Instruct	m2-ultra-192gb	68.4	FEEDER	69.1	76.4	68.7	68.7	67.7	55.0	5.7K	3262	2026-05-07
11	MiniCPM-V 2.6 8B	m3-pro-36gb	66.0	FEEDER	66.6	68.1	67.0	62.2	67.5	63.3	5.3K	822	2026-04-21
12	Hermes 3 Llama 3.1 8B	rtx-4080-16gb	64.7	FEEDER	63.2	64.7	68.7	58.2	67.4	68.1	5.4K	790	2026-05-14
13	Mistral 7B Instruct v0.3	ryzen-7950x-cpu-only	62.7	FEEDER	58.4	63.5	60.1	63.9	66.5	68.1	4.8K	775	2026-05-23
14	Phi 3.5 Mini	m2-air-16gb	59.4	TAP	62.5	56.9	57.4	53.5	61.1	64.7	5.3K	358	2026-04-27
15	StarCoder2 3B	ryzen-7950x-cpu-only	47.6	TAP	46.1	46.1	45.1	54.6	50.8	45.0	5.9K	343	2026-04-11
16	Llama 3.2 1B Instruct	m1-air-8gb	41.9	TAP	41.8	37.8	40.9	41.2	46.8	45.8	5.2K	329	2026-05-07
17	Qwen 2.5 0.5B Instruct	m1-air-8gb	32.2	DRIP	27.9	32.4	32.6	36.0	36.0	31.8	5.4K	360	2026-04-30