Datasets:

multimodalart
/

decision-index-leaderboard

rank int64 1 248	tied bool 2 classes	model stringlengths 3 43	engine stringlengths 3 35	organization stringlengths 5 24 ⌀	full_score float64 2.04 71.9	same_skills_eq float64 0.76 68.3	new_domains_eq float64 3.04 79.2	public_score float64 1.43 75.3	public_rank int64 1 248	params_b float64 0.02 36 ⌀	params_estimated bool 2 classes	kind stringclasses 6 values	technique stringclasses 6 values	base_model stringlengths 10 44 ⌀	model_url stringlengths 32 74 ⌀	code_url stringclasses 42 values	weights_repo stringclasses 38 values	revision stringclasses 23 values	weights_note stringclasses 1 value	latency_median_ms float64 0.3 524 ⌀	latency_mean_ms float64 0.7 1.21k ⌀	latency_p95_ms float64 1.7 3.91k ⌀	is_jev bool 2 classes	lab_run bool 1 class	new_in_edition bool 2 classes	submission_pr int64 21 331 ⌀
1	true	KnowLine 31B Gen1	knowline-31b-gen1	PelaAI	71.86	66.94	79.24	70.13	11	31.27	false	LoRA	autoregressive	google/gemma-4-31B-it	https://huggingface.co/PelaAI/KnowLine-31B-Gen1	null	null	null	null	203.7	459.3	1,822	false	false	true	283
2	true	Fastino GLiDE 1.2	glide-v1.2	Fastino	71.52	68.31	76.32	71.99	5	27.8	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/fastino/GLiDE	https://huggingface.co/fastino/GLiDE/tree/main/glide	fastino/GLiDE	a7e45896be7c96ba3354891da790c5ec99c30b6c	null	92.9	284	1,121.1	false	false	true	255
3	true	Sieve 31B v3	sieve-31b-v3	Sthānika AI	71.1	68.27	75.33	71.68	7	31.27	false	full fine-tune	autoregressive	google/gemma-4-31B-it	https://huggingface.co/sthanika-ai/Sieve-31B-v3	null	null	null	null	231.8	663	1,520.2	false	false	true	317
4	true	Turion Jev 31B	turion-jev-31b	Turion.AI	70.47	65.04	78.6	66.16	29	31.27	false	LoRA	autoregressive	google/gemma-4-31B-it	https://huggingface.co/turion-ai/turion-jev-31b	null	null	null	null	143.8	299.2	814.8	false	false	true	286
5	true	vLLM-SR d3-ultra (Decision 3.0, 27B)	vllm-sr-d3-ultra	vLLM Semantic Router	70.18	66.49	75.71	71.57	9	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/vllm-sr/d3-ultra	null	vllm-sr/d3-ultra	37fee01926f3de396620e794201b074fb427ac2d	null	200.5	423.7	673.5	false	false	true	302
6	true	Jeffan-31b-1710	jeffan-31b-1710	null	69.75	65.86	75.58	73.13	2	31.27	false	full fine-tune	autoregressive	google/gemma-4-31B-it	https://huggingface.co/tattrongvu/Jeffan-31b-1710	null	null	null	null	121	336.2	1,514	false	false	true	241
7	true	H2O Lightning 31B v1.1	h2o-lightning-31b-v1.1	H2O.ai	68.64	65.3	73.64	67.75	20	31.27	false	full fine-tune	autoregressive	google/gemma-4-31B-it	https://huggingface.co/h2oai/h2o-lightning-31b	null	null	null	null	100	207.1	701.6	false	false	true	276
8	true	WeMM-JEV-27B	wemm-jev-27b	null	68.49	65.09	73.6	69.37	14	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/Rover00/WeMM-JEV	null	null	null	null	98.3	279.5	1,076.1	false	false	true	273
9	true	baselabs mammoth-1 (27B)	mammoth-1	baselabs	68.44	65.98	72.14	75.34	1	27.36	false	full fine-tune	autoregressive	Qwen/Qwen3.5-27B	https://huggingface.co/baselabs/mammoth-1	null	baselabs/mammoth-1	9d1962f193e48a68cf0f3598dc7fb4462fd23ad9	null	220.1	435.1	712.5	false	false	true	260
10	true	Ainvest-Decision-27B v1.0	ainvest-decision-27b-v1.0	null	68.23	64.27	74.17	71.88	6	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/Ainvest-World/Ainvest-Decision-27b-v1.0	null	null	null	null	99.1	287.2	1,067.7	false	false	true	313
11	true	oenogarum (Gemma 4 31B)	oenogarum	null	68.1	62.5	76.5	62.8	48	31.27	false	full fine-tune	autoregressive	google/gemma-4-31B-it	https://huggingface.co/plantwaterer/oenogarum	null	null	null	null	259.4	616.7	1,989.7	false	false	true	292
12	true	Perplexity Decider v1.2 (27B)	pplx-decider-v1.2	Perplexity	67.98	63.99	73.98	67.74	22	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/perplexity-ai/pplx-decider-v1.2	null	null	null	null	144.4	415.1	1,552.2	false	false	false	null
13	true	AutoTrust-JEV-27B v1.1	autotrust-jev-27b-v11	AutoTrust AI Lab	67.95	63.77	74.22	67.32	24	27.36	false	full fine-tune	autoregressive	vllm-sr/d3	https://huggingface.co/autotrust/AutoTrust-JEV-27B-v1.1	null	null	null	null	79.9	224.7	861.2	false	false	true	231
14	true	StartLux-Decision v1.1 27B	startlux-decision-v1.1-27b-instant	StartLux	67.92	64.51	73.02	72.88	3	27.78	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/startlux-models/StartLux-Decision-v1.1-27B-Instant	null	null	null	null	79.8	265.9	853.5	false	false	false	null
15	true	MISA AI 2.0 Decision Plus · oak 1.1 (31B)	oak-1.1	MISA JSC	67.62	62.22	75.73	68.19	17	31.27	false	full fine-tune	autoregressive	google/gemma-4-31B-it	https://huggingface.co/PNPM244/oak1.1	null	null	null	null	112.6	407.4	1,611.7	false	false	true	295
16	true	AJev lm6q (Gemma 4 31B)	ajev-lm6q	null	67.57	63.05	74.34	65.46	34	31.27	false	LoRA	autoregressive	google/gemma-4-31B-it	https://huggingface.co/andyzhang232/ajev-gemma4-31b-lm6q	null	null	null	null	144.4	489	2,245.1	false	false	true	215
17	true	FlexDM · Gemma-4-31B	flexdm-gemma-4-31b	null	67.56	62.66	74.92	63.63	44	31.27	false	inference technique	autoregressive	google/gemma-4-31B-it	null	https://github.com/1000luca/FlexDM/tree/v1.0	null	null	null	108.6	300.9	1,106.1	false	false	true	246
18	true	Lychee-27B	lychee-27b	null	67.35	62.19	75.09	62.76	49	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/LycheeAI/Lychee-27B	null	null	null	null	97.2	320	1,158.4	false	false	false	null
19	true	vLLM-SR d3-turbo v1.1 (27B)	vllm-sr-d3-turbo-v11	vLLM Semantic Router	67.31	63.98	72.31	71.62	8	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/vllm-sr/d3-turbo	null	null	null	null	190.5	341.9	586.6	false	false	true	242
20	true	vLLM-SR d3 (27B)	vllm-sr-d3-v310	vLLM Semantic Router	67.15	62.5	74.11	65.45	35	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/vllm-sr/d3/tree/v3.1.0	null	null	null	null	97.1	318.9	1,144.3	false	false	false	null
21	true	allec-oavg (Gemma 4 31B)	allec-oavg	null	66.96	62.08	74.29	62.03	52	31.27	false	full fine-tune	autoregressive	google/gemma-4-31B-it	https://huggingface.co/plantwaterer/allec	null	null	null	null	256.8	613.5	1,977.6	false	false	true	287
22	true	VisionJEV 1.0.0 (27B)	visionjev-1.0.0	null	66.88	62.47	73.5	66.46	27	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/taoszhang/VisionJEV	null	null	null	null	89.9	324.8	1,099.7	false	false	true	239
23	true	Aleria Judge 1.1.1 (Gemma 4 31B)	aleria-judge-1.1.1	null	66.67	62.06	73.59	63.46	46	31.27	false	inference technique	autoregressive	google/gemma-4-31B-it	null	https://github.com/ALERIA-ML/aleria-judge/tree/v1.1.1	null	null	null	117.3	336.5	698	false	false	true	237
24	true	Nace.AI Drex Max v1 (27B)	drex-max-v1	Nace-Deepinfra	66.6	62.76	72.36	67.52	23	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	null	null	null	null	Open weights coming soon	103.6	325.3	1,158.6	false	false	true	238
25	true	Bespoke Nimble v4 Max (27B)	nimble-v4-max	Bespoke Labs	66.56	62.24	73.05	68.46	15	27.36	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/bespokelabs/Nimble-v4-max	null	null	null	null	107.1	253	632.5	false	false	true	226
26	true	AnyJev Tacitus 31B	anyjev-tacitus-31b	null	66.47	62.15	72.95	65.52	33	31.27	false	full fine-tune	autoregressive	google/gemma-4-31B-it	https://huggingface.co/morriszjm/AnyJev-Tacitus-31B	null	null	null	null	95.2	264.8	999.6	false	false	true	229
27	true	Decider 31B (NVFP4)	decider-31b	null	66.03	61.45	72.9	64.27	43	32.68	false	full fine-tune	autoregressive	google/gemma-4-31B-it	https://huggingface.co/Mapika/decider-31b	null	null	null	null	72.3	155.5	614.9	false	false	false	null
28	true	KnowLine 35B-A3B Gen1.5	knowline-35b-a3b-gen1.5	PelaAI	65.77	65.19	66.65	70.72	10	35.95	false	LoRA	autoregressive	Qwen/Qwen3.5-35B-A3B	https://huggingface.co/PelaAI/KnowLine-35B-A3B-Gen1.5	null	null	null	null	47.1	77.3	219.2	false	false	true	268
29	true	OneShot 27B	oneshot-27b	OneShot	65.74	61.71	71.78	65.41	36	27.78	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/OneShotOrg/OneShot	null	null	null	null	127.8	361	1,245.2	false	false	true	275
30	true	vera-large	vera-large	null	65.26	61.21	71.33	67.75	21	25.63	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/scar-ai/vera-large	null	null	null	null	159.7	407.4	1,265.2	false	false	true	281
31	true	Bobcat 1.3 (31B)	bobcat-1.3	null	64.85	59.54	72.8	59.31	67	31.27	false	LoRA	autoregressive	google/gemma-4-31B-it	https://huggingface.co/sanghwa-na/bobcat-1.3	null	null	null	null	63.8	182.7	767.1	false	false	false	null
32	true	Amos 1 Large (27B)	amos-1-large	null	64.81	62.37	68.48	64.94	41	27.78	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/mukaddamzaid/amos-1-large	null	null	null	null	147.9	460.7	1,365.8	false	false	false	null
33	true	Matilda Decision v1	matilda-decision-v1	Maincode	64.75	59.98	71.9	65.39	37	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/Maincode/matilda-decision-v1	null	null	null	null	91.9	308.2	1,150.1	false	false	true	189
34	true	jev-brain · Gemma-4-31B	jev-brain-gemma-4-31b	null	64.61	61.22	69.7	61.69	55	31.27	false	inference technique	autoregressive	google/gemma-4-31B-it	null	https://github.com/wcole3/jev-brain/tree/di-0.3-gemma-4-31b-r2	null	null	null	100.5	373.6	818.9	false	false	true	193
35	true	KnowLine 26B-A4B Gen2.5	knowline-26b-a4b-gen2.5	PelaAI	64.57	62.49	67.7	68.21	16	25.81	false	LoRA	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/PelaAI/KnowLine-26B-A4B-Gen2.5	null	null	null	null	71.7	124.9	417.5	false	false	true	271
36	true	frontier-infra Jebadiah 27B v2.1 (t1e)	jebadiah-27b-v2.1-t1e	Frontier Infra	64.4	58.98	72.52	60.46	60	27.78	false	LoRA	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/frontier-infra/jebadiah-27b-v2-1	null	null	null	null	303.2	809.5	2,950.2	false	false	true	199
37	true	AnyJev L0 · Gemma-4-31B	anyjev-l0-gemma-4-31b	null	63.65	58.76	70.99	59.79	63	31.27	false	inference technique	autoregressive	google/gemma-4-31B-it	null	https://github.com/MorrisZJ/anyjev-l0-gemma	null	null	null	97.9	260	1,080.3	false	false	false	null
38	true	deck31b (Gemma 4 31B)	deck31b	null	63.05	58.23	70.29	58.19	77	32.68	false	inference technique	autoregressive	google/gemma-4-31B-it	null	https://github.com/krishna-gogineni-765/deck31b	null	null	null	218.5	677.3	2,027.3	false	false	false	null
39	true	Darwin-27B-ZTC	darwin-27b-ztc	FINAL_Bench	63.03	58.34	70.07	62.63	50	25.62	false	full fine-tune	autoregressive	FINAL-Bench/Darwin-27B-RSI	https://huggingface.co/FINAL-Bench/Darwin-27B-ZTC	null	null	null	null	104	317.9	1,156.9	false	false	false	null
40	false	decisio 0.6.0 (Gemma 4 31B)	decisio-0.6.0-gemma-4-31b	null	62.68	58.12	69.52	58.09	78	31.27	false	inference technique	autoregressive	google/gemma-4-31B-it	null	https://github.com/aminry/decisio	null	null	null	53.9	151.6	475.2	false	false	false	null
41	true	Cerno 27B	cerno-27b	null	61.68	58.72	66.11	65.19	39	26.9	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/apertixai/cerno-27b	null	null	null	null	53.9	135.7	418	false	false	true	210
42	true	Decider chat · Gemma-4-31B	decider-chat-gemma4-31b	null	61.63	56.48	69.35	57.79	83	32.68	false	inference technique	autoregressive	google/gemma-4-31B	null	https://github.com/Mapika/decider	null	google/gemma-4-31B-it	null	108.5	358.6	1,114.7	false	false	false	null
43	true	Torchcast Decision 27B	torchcast-decision-27b	Torchcast AI	61.49	57.75	67.1	65.1	40	27.8	false	LoRA	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/torchcast-ai/torchcast-decision-27b	null	null	null	null	99.9	259	1,240	false	false	false	null
44	true	OneJev1.6-27B	onejev1.6-27b	OmniJev	61.46	62.15	60.43	69.55	13	27.36	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/OmniJev/OneJev1.6-27B	null	null	null	null	111.6	296.7	893.8	false	false	true	217
45	true	Qwen Decision 27B noncausal LoRA	qwen-decision-27b-noncausal	null	61.12	57.32	66.82	57.7	85	27.78	false	LoRA + head	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/jsaurabh/qwen-decision-27b-noncausal-lora	null	null	null	null	126.6	401.4	1,537.8	false	false	false	null
46	true	Quyet-1.0-Large	quyet-1.0-large	null	60.77	58.46	64.23	60.79	58	32.68	false	LoRA	autoregressive	google/gemma-4-31B-it	https://huggingface.co/chinhnc/Quyet-1.0-Large	null	null	null	null	105.4	356.6	1,588.6	false	false	false	null
47	true	garum (Gemma 4 31B)	garum	null	60.27	58.38	63.09	58.36	74	31.27	false	LoRA	autoregressive	google/gemma-4-31B-it	https://huggingface.co/plantwaterer/garum	null	null	null	null	119	450.4	1,705.3	false	false	true	184
48	true	baselabs elephant-1 (9B)	elephant-1-baselabs	baselabs	60.26	62.11	57.49	72.24	4	9.41	false	full fine-tune	autoregressive	Qwen/Qwen3.5-9B	https://huggingface.co/baselabs/elephant-1	null	baselabs/elephant-1	14537f08b0374edf9867ed0a4d2bd8df59e0400c	null	113.7	221.5	253.3	false	false	true	243
49	true	Jev 1.14	jev	null	60.24	58.91	62.24	60.7	59	null	null	null	null	null	null	null	null	null	null	null	null	null	true	false	false	null
50	true	Standard One 35B SH (v1)	standard-one-35b-sh-v1	Standard Thinking	60.14	61.43	58.2	66.26	28	35.11	false	LoRA + head	autoregressive	Qwen/Qwen3.6-35B-A3B	https://huggingface.co/StandardThinking/StandardOne-35B-SH	null	null	null	null	319.9	376.4	409.8	false	false	true	297
51	true	Lumma-fev-27B	lumma-fev-27b	FrontiersMind	59.97	56.2	65.64	57.15	92	26.09	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/FrontiersMind/Lumma-fev-27B	null	null	null	null	100	303.7	1,075.4	false	false	true	254
52	true	Sieve 27B v2	sieve-27b-v2	Sthānika AI	59.89	59.67	60.23	68.02	18	27.78	false	LoRA + head	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/sthanika-ai/Sieve-27B-v2	null	null	null	null	106.1	268.1	602.3	false	false	true	219
53	true	mogan-decision-31B-it	mogan-decision-31b-it	Mogan AI	59.82	57.28	63.64	65.55	32	31.27	false	LoRA	autoregressive	google/gemma-4-31B-it	https://huggingface.co/moganai/mogan-decision-31B-it	null	null	null	null	109	382.3	1,690.1	false	false	true	21
54	true	linne-fern 1.0 (27B)	linne-fern-1.0	MISA-AICORE	59.65	56.12	64.94	58.92	72	27.78	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/MISA-core/linne-fern-1-0	null	null	null	null	107.5	275.3	1,042.8	false	false	false	null
55	true	Kev 27B	kev-27b	null	59.56	55.1	66.26	56.69	94	27.8	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/jaredpalmer/kev-27b	null	null	null	null	135	227.8	null	false	false	false	null
56	true	Xor 26B-A4B v1.2	xor-26b-a4b-v1.2	Juspay Technologies	59.44	58.37	61.06	59.46	66	25.81	false	LoRA	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/juspay/Xor-26B-A4B	null	null	null	null	109.9	168.1	479.8	false	false	true	202
57	true	Jev 1.13	jev-1.13.0	null	59.22	57.66	61.58	57.96	82	null	false	hosted API	null	null	null	null	null	null	null	524	524	536.1	false	false	false	null
58	true	span-02 (Gemma 4 26B-A4B)	span-02	null	59.07	58.95	59.26	63.62	45	25.81	false	full fine-tune	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/lihe07/span-02	null	null	null	null	33	234.1	822.4	false	false	false	null
59	true	garas (Gemma 4 31B)	garas	null	58.95	56.24	63.03	56.59	95	31.27	false	LoRA	autoregressive	google/gemma-4-31B-it	https://huggingface.co/plantwaterer/garas	null	null	null	null	117.1	488.5	1,858.4	false	false	false	null
60	true	Shisa DE-2.1 (repeat)	shisa-de-2.1-repeat	Shisa.AI	58.84	56.22	62.76	59.23	69	25.81	false	LoRA	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/shisa-ai/shisa-de-2.1	null	null	null	null	54.5	159.3	512.6	false	false	true	236
61	true	Eikos-27B-FP8 Qwen3.8-27B LoRA	eikos-27b-fp8	null	58.79	55.52	63.69	52.85	109	27.78	false	LoRA	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/caiovicentino1/Eikos-27B-FP8	https://github.com/caiovicentino/eikos	caiovicentino1/Eikos-27B-FP8	null	null	129.7	213.2	509.8	false	false	false	null
62	true	reflex Qwen3.8-27B-FP8 (pmscr)	reflex-27b-pmscr	null	58.77	57	61.43	57.34	90	27.78	false	inference technique	autoregressive	Qwen/Qwen3.8-27B-FP8	null	https://github.com/kshetrajna12/reflex	null	null	null	148.5	318.7	1,251.3	false	false	false	null
63	true	RSI-Jev v6.1 VL 27B (effort auto)	rsi-jev-v6.1-vl-27b-auto	null	58.6	57.81	59.79	65.64	31	27.78	false	LoRA	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/shgao/rsi-jev-v6.1-vl-27b	null	null	null	null	76.1	182.5	564.5	false	false	false	null
64	true	KnowLine 9B Gen1	knowline-9b-gen1	PelaAI	58.44	60.48	55.38	65.67	30	9.65	false	LoRA	autoregressive	Qwen/Qwen3.5-9B	https://huggingface.co/PelaAI/KnowLine-9B-Gen1	null	null	null	null	128.1	222.6	747.1	false	false	true	289
65	true	StartLux-Decision v1.1 9B	startlux-decision-v1.1-9b-instant	StartLux	58.07	60.95	53.75	69.98	12	9.65	false	full fine-tune	autoregressive	Qwen/Qwen3.5-9B	https://huggingface.co/startlux-models/StartLux-Decision-v1.1-9B-Instant	null	null	null	null	38.6	93	261.4	false	false	true	208
66	true	vLLM-SR Decision 2.0 Vega 27B	decision-2.0-vega-27b	vLLM Semantic Router	58.01	55.3	62.07	56.97	93	27.8	false	LoRA + head	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/vllm-sr/Decision-2.0-Vega-27B	null	null	null	null	119.6	239.2	null	false	false	false	null
67	true	JEV-35B	jev-35b	AutoTrust AI Lab	57.61	55.78	60.36	59.73	64	35.95	false	LoRA + head	autoregressive	Qwen/Qwen3.5-35B-A3B	https://huggingface.co/autotrust/JEV-35B	null	null	null	null	50.5	90.2	258.3	false	false	false	null
68	true	JEV-27B	jev-27b	AutoTrust AI Lab	57.35	53.79	62.67	53.64	105	27.8	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/autotrust/JEV-27B	null	null	null	null	102.1	584.6	null	false	false	false	null
69	true	simple-jev · Qwen3.8-27B (featherless)	simple-jev-qwen3.8-27b	null	56.99	55.05	59.9	56.05	98	27.78	false	inference technique	autoregressive	Qwen/Qwen3.8-27B	null	https://github.com/featherless-ai/simple-jev	null	Qwen/Qwen3.8-27B	null	373	930.6	2,796.8	false	false	false	null
70	true	Tern-35B-A3B	tern-35b-a3b	null	56.62	55.59	58.16	59.01	71	34.77	false	full fine-tune	autoregressive	Qwen/Qwen3.6-35B-A3B	https://huggingface.co/matthewagi/Tern-35B-A3B	null	null	null	null	62.9	82.8	139.2	false	false	true	227
71	true	Surogate Rune 26B-A4B v3	rune-26b-a4b-v3	Surogate	56.42	56.7	56	58.28	75	25.81	false	full fine-tune	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/surogate/rune-26b-a4b-GGUF	https://github.com/invergent-ai/surogate	surogate/rune-26b-a4b-GGUF	bd4a7cbbed66af1a95dfbac12cabf52e6e677411	null	120.5	267.2	682.8	false	false	false	null
72	true	SynACK Decide 26B-A4B v7	synack-decide-26b-a4b-v7	null	56.22	55.89	56.72	59.68	65	25.81	false	LoRA	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/SkyPanther/synack-decide-26b-a4b	null	null	null	null	129.1	174.1	363.7	false	false	false	null
73	true	KnowLine 4B Gen5.5	knowline-4b-gen5.5	PelaAI	55.99	60.85	48.69	66.76	26	4.66	false	full fine-tune	autoregressive	Qwen/Qwen3.5-4B	https://huggingface.co/PelaAI/KnowLine-4B-Gen5.5	null	null	null	null	35.2	61.1	170.3	false	false	true	272
74	true	Blink v0.3 26B-A4B (NVFP4)	blink-v0.3-26b-a4b-nvfp4	Pixilab.ai	55.92	56.81	54.57	58.21	76	25.8	false	full fine-tune	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/PixilabAI/Blink-v0.3-26B-A4B-NVFP4	null	null	null	null	25.1	48.4	null	false	false	false	null
75	true	JADE	jade	null	55.57	53.86	58.12	53.42	107	27.8	false	LoRA	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/theunnecessarythings/JADE	null	null	null	null	133.3	277.7	null	false	false	false	null
76	true	Wald-26B-A4B	wald-26b-a4b	ORG2 AI	55.35	55.98	54.41	62.3	51	25.81	false	full fine-tune	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/org2ai/Wald-26B-A4B	null	null	null	null	90	167.8	339.9	false	false	false	null
77	true	AJev lora1 (Gemma 4 26B-A4B)	ajev-gemma26-lora1	null	55.29	54.5	56.47	57.67	86	25.81	false	LoRA	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/andyzhang232/ajev-gemma4-26b-a4b-lora1	null	null	null	null	88.8	94.3	247	false	false	false	null
78	true	Cloudflare clef	clef	Cloudflare	54.92	52.67	58.3	61.71	54	27.8	false	full fine-tune	autoregressive	Qwen/Qwen3.8-27B	https://huggingface.co/Cloudflare/clef	null	null	null	null	102.1	245.5	null	false	false	false	null
79	true	StartLux-Decision v1.1 4B	startlux-decision-v1.1-4b-instant	StartLux	54.8	58.22	49.67	67.81	19	4.66	false	full fine-tune	autoregressive	Qwen/Qwen3.5-4B	https://huggingface.co/startlux-models/StartLux-Decision-v1.1-4B-Instant	null	null	null	null	38.3	68.8	165.7	false	false	true	208
80	true	Decider chat · Qwen3.6-27B	decider-chat-qwen3.6-27b	null	54.67	51.7	59.11	51.57	113	27.78	false	inference technique	autoregressive	Qwen/Qwen3.6-27B	null	https://github.com/Mapika/decider	null	Qwen/Qwen3.6-27B	null	83.6	284	965.6	false	false	false	null
81	true	Perplexity Decider Flash v1.2 (4B)	pplx-decider-flash-v1.2	Perplexity	54.29	55.76	52.07	60.06	61	4.54	false	full fine-tune	autoregressive	Qwen/Qwen3.5-4B	https://huggingface.co/perplexity-ai/pplx-decider-flash-v1.2	null	null	null	null	74.8	136.8	380.5	false	false	false	null
82	true	vLLM-SR d3-flash (9B)	vllm-sr-d3-flash	vLLM Semantic Router	54.27	54.21	54.36	57.79	84	8.39	false	full fine-tune	autoregressive	Qwen/Qwen3.5-9B	https://huggingface.co/vllm-sr/d3-flash	null	null	null	null	53.7	113.1	353.2	false	false	false	null
83	true	JPT-35B-A3B	jpt-35b-a3b	null	54.09	51.8	57.52	52.68	110	35.95	true	LoRA	autoregressive	Qwen/Qwen3.5-35B-A3B	https://huggingface.co/kirp/jpt-35b-a3b	null	null	null	null	250.2	339.7	794.6	false	false	false	null
84	true	Xor 1.2	xor-1.2	Juspay Technologies	53.62	53.17	54.29	52.28	111	35.95	false	LoRA	autoregressive	Qwen/Qwen3.6-35B-A3B	https://huggingface.co/juspay/xor	null	null	null	null	149.9	186.5	null	false	false	false	null
85	true	GEV-26B-Decide	jev-gemma4-26b-a4b	AutoTrust AI Lab	53.5	54.58	51.88	58.03	81	25.8	false	LoRA	autoregressive	google/gemma-4-26B-A4B-it	https://huggingface.co/autotrust/GEV-26B-Decide	null	null	null	null	128.5	228.3	null	false	false	false	null
86	true	OneJev1.6-9B	onejev1.6-9b	OmniJev	53.05	56.46	47.93	65.3	38	9.41	false	full fine-tune	autoregressive	Qwen/Qwen3.5-9B	https://huggingface.co/OmniJev/OneJev1.6-9B	null	null	null	null	42.2	120	305.4	false	false	true	217
87	true	Nace.AI Drex v2.0 (35B-A3B)	drex-v2.0	Nace.AI	52.76	53.87	51.08	61.25	56	35.11	false	full fine-tune	autoregressive	Qwen/Qwen3.6-35B-A3B	https://huggingface.co/nace-ai/drex-v2.0	null	null	null	null	275.9	329.7	529.3	false	false	false	null
88	true	muria (Gemma 4 12B)	muria	null	52.55	53.07	51.78	51.4	115	11.96	false	full fine-tune	autoregressive	google/gemma-4-12B-it	https://huggingface.co/plantwaterer/muria	null	null	null	null	53.5	183.1	698.9	false	false	true	256
89	true	Jeffan-4B-2010	jeffan-4b-2010	null	52.21	57.65	44.04	64.85	42	4.21	false	full fine-tune	autoregressive	Qwen/Qwen3.5-4B	https://huggingface.co/tattrongvu/Jeffan-4B-2010	null	null	null	null	50.4	87.4	236.1	false	false	true	259
90	false	Verdin-4B v1.1	verdin-4b-v1.1	null	51.99	53.7	49.43	60.96	57	4.21	false	full fine-tune	autoregressive	Qwen/Qwen3.5-4B	https://huggingface.co/victorbennett/Verdin-4B-v1.1	null	null	null	null	199.9	294.2	505.4	false	false	true	300
91	true	Bespoke Nimble 9B v3	nimble-v3	Bespoke Labs	51.06	51.89	49.82	57.19	91	9.65	false	LoRA	autoregressive	Qwen/Qwen3.5-9B	https://huggingface.co/bespokelabs/Bespoke-Nimble-9B-v3	null	null	null	null	73.9	251.2	null	false	false	false	null
92	true	Senova-4B v2	senova-4b-v2	null	51.01	55.38	44.45	66.87	25	4.54	false	full fine-tune	autoregressive	Qwen/Qwen3.5-4B	https://huggingface.co/WangSenQWQ/Senova-4B-v2	null	null	null	null	53.1	177.6	471.7	false	false	true	324
93	true	frontier-infra Jebadiah 9B v2.1 (t1e)	jebadiah-9b-v2.1-t1e	Frontier Infra	50.9	50.57	51.39	51.53	114	9.65	false	LoRA	autoregressive	Qwen/Qwen3.5-9B	https://huggingface.co/frontier-infra/jebadiah-9b-v2-1	null	null	null	null	125.1	294.2	968	false	false	true	188
94	true	decisio 0.4.0 (Qwen3.6-35B-A3B)	decisio-0.4.0-qwen3.6-35b-a3b	null	50.82	48.67	54.06	48.43	129	35.95	false	inference technique	autoregressive	Qwen/Qwen3.6-35B-A3B	null	https://github.com/aminry/decisio	null	null	null	70.5	112.9	261.2	false	false	false	null
95	true	vLLM-SR d3-mini (4B)	vllm-sr-d3-mini	vLLM Semantic Router	49.97	51.45	47.75	54.9	102	4.54	false	full fine-tune	autoregressive	Qwen/Qwen3.5-4B	https://huggingface.co/vllm-sr/d3-mini	null	null	null	null	50.4	87.2	237.1	false	false	false	null
96	true	JoshuaSP diffgemma	joshua-diffusion-full	null	49.53	49.54	49.5	50.14	122	25.82	false	inference technique	diffusion	google/diffusiongemma-26B-A4B-it	null	https://github.com/JoshuaSP/open-jev	null	google/diffusiongemma-26B-A4B-it	null	260.4	337.8	537.3	false	false	false	null
97	true	AJev lora7 (Gemma 4 12B)	ajev-lora7	null	48.91	50.04	47.2	55.17	101	11.96	false	LoRA	autoregressive	google/gemma-4-12B-it	https://huggingface.co/andyzhang232/ajev-gemma4-12b-lora7	null	null	null	null	69.1	117.2	453	false	false	false	null
98	true	OneJev1.6-4B	onejev1.6-4b	OmniJev	48.88	55.37	39.15	61.95	53	4.54	false	full fine-tune	autoregressive	Qwen/Qwen3.5-4B	https://huggingface.co/OmniJev/OneJev1.6-4B	null	null	null	null	27.2	90.4	201.1	false	false	true	217
99	true	decisio 0.4.0 (Gemma 4 12B)	decisio-0.4.0-gemma-4-12b	null	48.75	47.94	49.95	49.72	126	11.96	false	inference technique	autoregressive	google/gemma-4-12B-it	null	https://github.com/aminry/decisio	null	null	null	37.8	111.9	275.9	false	false	false	null
100	true	Mica v0.5 4B	mica-v0.5-4b	akivet	47.7	51.81	41.55	57.56	88	4.66	false	full fine-tune	autoregressive	Qwen/Qwen3.5-4B	https://huggingface.co/sky7350/Mica-v0.5-4B	https://github.com/akivet/Mica/tree/4adbdd398d6e2a9ca0a63b1e3d3b49240f03cee5	null	null	null	19.7	70.5	147.5	false	false	true	258

End of preview. Expand in Data Studio

Decision Index

The text board of the Jev Decision Index (Decision Index 0.4.1): models that read a state and a typed question and return a decision. Vision board: multimodalart/decision-index-vision-leaderboard.

Tasks (each is its own leaderboard):

decision_index: the Decision Index (Full) = 0.6 × same-skills + 0.4 × new-domains, both equated on 21 stock base models. Scores within 0.9 points share a rank.
same_skills: private sets that test the same skills as the public suite.
new_domains: private sets from domains outside the public suite.
public: the 42-benchmark public suite.

The private and new-domain sets are not released; only aggregate scores are published. The evaluation kit and submission process: apolinario/decision-index. Methodology: https://huggingface.co/spaces/multimodalart/jev-decision-index.

This repo mirrors the board: leaderboard has one row per model (248 entries), benchmarks the per-benchmark public scores. It is regenerated whenever the board is updated.

Top 10

Rank	Model	Full
1	KnowLine 31B Gen1	71.86
2	Fastino GLiDE 1.2	71.52
3	Sieve 31B v3	71.10
4	Turion Jev 31B	70.47
5	vLLM-SR d3-ultra (Decision 3.0, 27B)	70.18
6	Jeffan-31b-1710	69.75
7	H2O Lightning 31B v1.1	68.64
8	WeMM-JEV-27B	68.49
9	baselabs mammoth-1 (27B)	68.44
10	Ainvest-Decision-27B v1.0	68.23

Generated 2026-10-29T14:04:16Z.

Downloads last month: 392

Decision Index

Top 10

Space using multimodalart/decision-index-leaderboard 1