General

When2Call (Decision Index) leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap
License: https://creativecommons.org/licenses/by-nc/4.0/

As of 2026-10-29, the highest measured score on When2Call (Decision Index) is 94.8% by elephant-1. 88 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	mammoth-1	100.0%	estimated ± 3.5 pp, low confidence
2	elephant-1	94.8%	measured
3	d1-ultra	92.6%	measured
4	Drex v2.0	89.7%	estimated ± 3.5 pp, medium confidence
5	Perplexity Decider v1.2 27B	87.7%	measured
6	Blink v0.3 26B-A4B (NVFP4)	85.1%	estimated ± 3.5 pp, medium confidence
7	Torchcast Decision 27B	84.9%	estimated ± 3.5 pp, medium confidence
8	Surogate Rune 26B-A4B v3	84.6%	estimated ± 3.5 pp, medium confidence
9	Celeris-1 Decision	83.7%	measured
10	GLiDE	83.7%	measured
11	Microsoft-Decision-1	81.9%	measured
12	Drex v1.5	81.6%	estimated ± 3.5 pp, medium confidence
13	Jev 1.14	80.8%	measured
14	Mercury Decide	79.3%	measured
15	GEV-26B-Decide	79.3%	estimated ± 3.5 pp, medium confidence
16	Xor 1.2	78.5%	estimated ± 3.5 pp, medium confidence
17	d1	77.4%	measured
18	Bespoke Nimble 9B v3	77.1%	estimated ± 3.5 pp, medium confidence
19	Eikos-27B	76.9%	estimated ± 3.5 pp, medium confidence
20	JEV-27B	76.6%	estimated ± 3.5 pp, medium confidence
21	vLLM-SR Decision 2.0 Vega 27B	74.6%	estimated ± 3.5 pp, medium confidence
22	Perplexity Decider Flash v1.2 4B	74.4%	measured
23	simple-jev · Qwen3.8-27B	74.1%	estimated ± 3.5 pp, medium confidence
24	JADE	73.9%	estimated ± 3.5 pp, medium confidence
25	Kev 27B	73.8%	estimated ± 3.5 pp, medium confidence
26	Quyet-1.0-Large	73.8%	estimated ± 3.5 pp, medium confidence
27	deck31b (Gemma 4 31B)	72.5%	estimated ± 3.5 pp, medium confidence
28	Decider chat · Gemma-4-31B	72.5%	estimated ± 3.5 pp, medium confidence
29	Clef	71.0%	measured
30	decider-35b-a3b	70.3%	estimated ± 3.5 pp, medium confidence
31	Decider chat · Qwen3.6-27B	70.3%	estimated ± 3.5 pp, medium confidence
32	JPT-35B-A3B	68.0%	estimated ± 3.5 pp, medium confidence
33	JoshuaSP diffusiongemma (open-jev)	67.2%	estimated ± 3.5 pp, medium confidence
34	decisio 0.4.0 (Qwen3.6-35B-A3B)	66.1%	estimated ± 3.5 pp, medium confidence
35	lev	64.8%	estimated ± 3.5 pp, medium confidence
36	Clef-Flash	64.4%	measured
37	Bosun v3.1 1.7B	63.8%	estimated ± 8.9 pp, low confidence
38	Kodiak v0.2 1B (accuracy)	63.8%	estimated ± 8.9 pp, low confidence
39	Bekko System One v0 68M	63.8%	estimated ± 8.9 pp, low confidence
40	Bosun v3.1 0.6B	63.8%	estimated ± 8.9 pp, low confidence
41	vLLM-SR Decision 2.0 Kai 0.6B	63.8%	estimated ± 8.9 pp, low confidence
42	Sifr 0.8B v3.1	63.8%	estimated ± 8.9 pp, low confidence
43	Jev-Style 0.8B v3	63.8%	estimated ± 8.9 pp, low confidence
44	Von 1.3	63.8%	estimated ± 8.9 pp, low confidence
45	JevK5 Lite	63.8%	estimated ± 8.9 pp, low confidence
46	GLiNER2.5-Decide	63.8%	estimated ± 8.9 pp, low confidence
47	jeff	63.8%	estimated ± 8.9 pp, low confidence
48	Lavoir	63.8%	estimated ± 8.9 pp, low confidence
49	Kev 0.8B r15	63.8%	estimated ± 8.9 pp, low confidence
50	Bekko System One v0 17M	63.8%	estimated ± 8.9 pp, low confidence
51	decisio 0.4.0 (Gemma 4 12B)	63.7%	estimated ± 3.5 pp, medium confidence
52	Lev1 3.7B	63.7%	estimated ± 3.5 pp, medium confidence
53	ezjev 4B s2	62.3%	estimated ± 3.5 pp, medium confidence
54	Clef-Omni	62.1%	measured
55	Jevfire	62.0%	estimated ± 3.5 pp, low confidence
56	vLLM-SR Decision 2.0 Lux 9B	62.0%	estimated ± 3.5 pp, low confidence
57	djev	61.7%	estimated ± 3.5 pp, low confidence
58	Hopper 12B (trained)	61.5%	estimated ± 3.5 pp, low confidence
59	razorback16 openjev diffusiongemma (NVFP4, vLLM)	59.3%	estimated ± 3.5 pp, low confidence
60	Ryotide Qwen v0.3.1	59.0%	estimated ± 3.5 pp, low confidence
61	JevK5 9B v0.3.3	58.1%	estimated ± 3.5 pp, low confidence
62	Jobe	58.1%	estimated ± 3.5 pp, low confidence
63	Winnow-12B Q8	58.1%	estimated ± 3.5 pp, low confidence
64	JEV-9B	57.3%	estimated ± 3.5 pp, low confidence
65	Decider 4B	57.1%	estimated ± 3.5 pp, low confidence
66	Bespoke Nimble 9B v2	56.6%	estimated ± 3.5 pp, low confidence
67	Hopper (G) 1.2	56.3%	estimated ± 3.5 pp, low confidence
68	JevK5 v0.3	55.5%	estimated ± 3.5 pp, low confidence
69	JPT-9B	55.1%	estimated ± 3.5 pp, low confidence
70	mmastrac diffusiongemma (vLLM PR 57250)	55.1%	estimated ± 3.5 pp, low confidence
71	SemIf	54.8%	estimated ± 3.5 pp, low confidence
72	Intern-Decision-4B	54.6%	estimated ± 3.5 pp, low confidence
73	openvons	54.3%	estimated ± 3.5 pp, low confidence
74	mini-jev	53.0%	estimated ± 3.5 pp, low confidence
75	vLLM-SR Decision 2.0 Nox 4B	53.0%	estimated ± 3.5 pp, low confidence
76	Jev-Omni	53.0%	estimated ± 3.5 pp, low confidence
77	Kas 4B	52.6%	estimated ± 3.5 pp, low confidence
78	Jevstral 8B	52.3%	estimated ± 5.9 pp, low confidence
79	Solomon v1.1	51.1%	estimated ± 3.5 pp, low confidence
80	JPT-4B	51.0%	estimated ± 3.5 pp, low confidence
81	NeoHorse-Jev-4B	49.9%	estimated ± 5.9 pp, low confidence
82	Sieve 9B	49.6%	estimated ± 3.5 pp, low confidence
83	Gevva E4B	47.4%	estimated ± 3.5 pp, low confidence
84	Winnow-E4B	47.2%	estimated ± 3.5 pp, low confidence
85	Tev1-4B-experimental	45.8%	estimated ± 3.5 pp, low confidence
86	Intelif Qwen3-4B	45.6%	estimated ± 3.5 pp, low confidence
87	Bekko System One v0 400M	41.5%	estimated ± 3.5 pp, low confidence
88	decider-2b	41.2%	estimated ± 3.5 pp, low confidence
89	Sieve 2B	40.0%	estimated ± 3.5 pp, low confidence
90	vLLM-SR Decision 2.0 Sol 2B	39.3%	estimated ± 5.9 pp, low confidence
91	Kev 4B r10	39.0%	estimated ± 3.5 pp, low confidence
92	metask-jev-4b	37.8%	estimated ± 3.5 pp, low confidence
93	RSI-Jev v5.0 VL 3B	37.6%	estimated ± 3.5 pp, low confidence
94	Gevva E2B	37.5%	estimated ± 5.9 pp, low confidence
95	vLLM-SR Decision 2.0 Eos 0.8B	37.4%	estimated ± 6.8 pp, low confidence
96	EXAONE-4.0-1.2B-JEV v0.3	36.8%	estimated ± 5.9 pp, low confidence
97	JPT-0.8B	36.7%	estimated ± 6.8 pp, low confidence
98	Intern-Decision-2B	36.5%	estimated ± 3.5 pp, low confidence
99	Kev 9B v2	36.4%	estimated ± 3.5 pp, low confidence
100	open-jev (pngwn)	34.8%	estimated ± 3.5 pp, low confidence
101	this-that 1.2	32.9%	estimated ± 3.5 pp, low confidence

When2Call (Decision Index) leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General