Decision Index 0.3 Retrieval

We show this table for reference; we do not rank on it.

Data verified October 6, 2026

Decision Index 0.3 publishes chance-corrected retrieval scores for 112 evaluated systems.

Retrieval score on Decision Index 0.3 Retrieval — October 6, 2026

We mirror the published retrieval score view for Decision Index 0.3 Retrieval. The public snapshot contains 112 evaluated systems. Perplexity Decider v1.1 (27B) has the highest published score at 64.7 points, but the available coverage and evaluation setups do not establish a market leader. We do not use these results to rank models overall.

1Open weights

Perplexity Decider v1.1 (27B)

Perplexity

64.7 points

2Open weights

GEV-26B-Decide

autotrust

64.6 points

3Evaluated configuration

Torchcast Decision 27B

Not listed

64.5 points

112 evaluated systemsDecision ModelsCurrentDisplay onlyUpdated October 6, 2026

Retrieval score table (112 evaluated systems)

Showing 112 of 112

Retrieval score table with model size, weight access, and scores
Rank	Model / configuration	Reported size	Weights	Score
1	Perplexity Decider v1.1 (27B)PerplexityModel sourceAnswered 100% · Latency 0.1s	28B	Open	64.7 points
2	GEV-26B-DecideautotrustModel sourceAnswered 100% · Latency 0.1s	26B	Open	64.6 points
3	Torchcast Decision 27BNot listedAnswered 100% · Latency 0.1s	28B	Evaluated configuration	64.5 points
4	deck31b (Gemma 4 31B)Not listedAnswered 91.5% · Latency 0.2s	33B	Evaluated configuration	64.2 points
5	Quyet-1.0-LargechinhncModel sourceAnswered 91.5% · Latency 0.1s	33B	Open	64.0 points
6	simple-jev · Qwen3.8-27B (featherless)Featherless AIModel sourceCalibration 11.3 · Answered 99.6% · Latency 0.4s	28B	Evaluated configuration	63.4 points
7	Decider chat · Gemma-4-31BMapikaModel sourceCalibration 4.7 · Answered 99.6% · Latency 0.1s	33B	Evaluated configuration	63.3 points
8	Blink v0.3 26B-A4B (NVFP4)PixilabAIModel sourceAnswered 100% · Latency 0.0s	26B	Open	63.1 points
9	Surogate Rune 26B-A4B v3SurogateModel sourceCalibration 12 · Answered 99.6% · Latency 0.1s	26B	Open	62.7 points
10	JevNot listedCalibration 7.4 · Answered 100% · Latency 0.5s	Not disclosed	Evaluated configuration	61.4 points
11	Xor 26B-A4BjuspayModel sourceAnswered 100% · Latency 0.1s	26B	Open	61.1 points
12	vLLM-SR Decision 2.0 Vega 27BNot listedAnswered 100% · Latency 0.1s	28B	Evaluated configuration	61.0 points
13	Cloudflare clefCloudflareModel sourceCalibration 4 · Answered 100% · Latency 0.1s	28B	Open	60.7 points
14	Fastino GLiDE no-thinking (28B)Not listedCalibration 4 · Answered 100% · Latency 0.1s	28B	Evaluated configuration	60.0 points
15	Eikos-27Bcaiovicentino1Model sourceCalibration 5.7 · Answered 100% · Latency 0.1s	28B	Not listed	58.3 points
16	reflex 27Bkshetrajna12Calibration 2.4 · Answered 100% · Latency 0.1s	28B	Open	58.1 points
17	JEV-27BautotrustModel sourceAnswered 100% · Latency 0.1s	28B	Open	57.1 points
18	Xor 1.2Not listedAnswered 100% · Latency 0.1s	36B	Evaluated configuration	56.9 points
19	decisio 0.4.0 (Gemma 4 12B)Not listedAnswered 100% · Latency 0.0s	12B	Evaluated configuration	56.9 points
20	Jebadiah 27Bfrontier-infraModel sourceCalibration 1.4 · Answered 100% · Latency 0.1s	28B	Open	55.8 points
21	Kev 27BjaredpalmerModel sourceCalibration 2.2 · Answered 100% · Latency 0.1s	28B	Open	55.6 points
22	JADENot listedAnswered 100% · Latency 0.1s	28B	Evaluated configuration	55.5 points
23	JoshuaSP diffusiongemma (open-jev)Not listedCalibration 21.6 · Answered 99.7% · Latency 0.3s	26B	Evaluated configuration	55.4 points
24	AJev lora5 (Gemma 4 12B)andyzhang232Model sourceAnswered 100% · Latency 0.3s	12B	Open	55.3 points
25	JPT-35B-A3BkirpModel sourceAnswered 91.5% · Latency 0.3s	~36B	Open	55.2 points
26	vLLM-SR Decision 2.0 Lux 9BNot listedAnswered 100% · Latency 0.0s	9.7B	Evaluated configuration	54.7 points
27	Winnow-12BEldan RingModel sourceCalibration 16.8 · Answered 100% · Latency 0.1s	12B	Open	54.4 points
28	JevfireNot listedCalibration 5.2 · Answered 100% · Latency 0.1s	28B	Evaluated configuration	54.4 points
29	decisio 0.4.0 (Qwen3.6-35B-A3B)Not listedAnswered 100% · Latency 0.1s	36B	Evaluated configuration	53.9 points
30	Bespoke Nimble 9B v3bespokelabsModel sourceAnswered 100% · Latency 0.1s	9.7B	Open	53.6 points
31	Decider chat · Qwen3.6-27BNot listedCalibration 2.1 · Answered 99.6% · Latency 0.1s	28B	Evaluated configuration	53.3 points
32	ezjev 4B s2everettjfModel sourceAnswered 100% · Latency 0.0s	4.7B	Open	53.2 points
33	Hopper 12B (trained)Not listedAnswered 91.5% · Latency 0.3s	12B	Evaluated configuration	53.0 points
34	Decider 35B-A3BMapikaModel sourceCalibration 2.3 · Answered 100% · Latency 0.1s	36B	Open	52.9 points
35	vLLM-SR Decision 2.0 Nox 4BNot listedAnswered 100% · Latency 0.0s	4.2B	Evaluated configuration	52.3 points
36	JevK5 9B v0.3.3alibiserikbayModel sourceAnswered 100% · Latency 0.0s	9.7B	Open	51.5 points
37	jiwo 4BeljiwoModel sourceAnswered 100% · Latency 0.1s	4.7B	Open	51.3 points
38	Jev-Style 2B v3Not listedAnswered 100% · Latency 0.1s	2.3B	Evaluated configuration	50.1 points
39	JEV-9BautotrustModel sourceAnswered 100% · Latency 0.1s	9.7B	Open	47.9 points
40	Jet v6.2michaljachModel sourceCalibration 14.1 · Answered 99.8% · Latency 0.0s	4.7B	Open	47.8 points
41	Kev 9B v2Not listedCalibration 4.5 · Answered 100% · Latency 0.1s	9.7B	Evaluated configuration	47.3 points
42	RSI-Jev v5.0 VL 3BshgaoModel sourceAnswered 100% · Latency 0.0s	4.7B	Open	46.8 points
43	Ryotide Qwen v0.3.1Not listedAnswered 100% · Latency 0.1s	4.0B	Evaluated configuration	46.3 points
44	Hopper (G) 1.2HopitAIModel sourceCalibration 9.3 · Answered 99.6% · Latency 0.0s	4.7B	Open	46.2 points
45	Sieve 9Bsthanika-aiModel sourceAnswered 100% · Latency 0.1s	9.7B	Open	45.9 points
46	JevK5 v0.3Not listedAnswered 100% · Latency 0.0s	4.7B	Evaluated configuration	45.3 points
47	Lev1 3.7BNot listedAnswered 100% · Latency 0.0s	~3.7B	Evaluated configuration	44.9 points
48	NeoHorse-Jev-4BTokenRhythmModel sourceCalibration 10.4 · Answered 99.6% · Latency 0.0s	4.7B	Open	44.4 points
49	Cloudflare clef-flashCloudflareModel sourceCalibration 2.5 · Answered 100% · Latency 0.1s	9.7B	Open	44.1 points
50	djevMaisaCalibration 21.2 · Answered 98.9% · Latency 0.1s	26B	Open	44.0 points
51	levinterfaze-aiModel sourceCalibration 6.4 · Answered 99.6% · Latency 0.1s	4.7B	Open	43.9 points
52	JPT-9BkirpModel sourceCalibration 7.7 · Answered 100% · Latency 0.1s	9.7B	Open	43.0 points
53	jiwo 0.8BeljiwoModel sourceAnswered 100% · Latency 0.0s	870M	Open	43.0 points
54	Decider 4BMapikaModel sourceCalibration 8.4 · Answered 99.6% · Latency 0.0s	4.7B	Open	42.9 points
55	Kev 4B r10jaredpalmerModel sourceCalibration 3.2 · Answered 100% · Latency 0.1s	4.7B	Open	41.9 points
56	vLLM-SR Decision 2.0 Sol 2BNot listedAnswered 100% · Latency 0.0s	2.3B	Evaluated configuration	41.9 points
57	JPT-4BkirpModel sourceCalibration 7.9 · Answered 100% · Latency 0.1s	4.7B	Open	41.7 points
58	Winnow-E4BEldanRingModel sourceCalibration 5.8 · Answered 99.6% · Latency 0.0s	8.0B	Open	41.6 points
59	Kas 4BkpiyaModel sourceAnswered 91.5% · Latency 0.1s	4.0B	Open	40.7 points
60	Intern-Decision-4BinternlmModel sourceCalibration 2.8 · Answered 99.6% · Latency 0.0s	4.7B	Open	39.8 points
61	Bosun v3.1 1.7BHanno-LabsModel sourceCalibration 13 · Answered 99.6% · Latency 0.0s	1.7B	Open	39.3 points
62	Bekko System One v0 400MhotchpotchModel sourceAnswered 100% · Latency 0.0s	~400M	Open	39.0 points
63	Gevva E2BdavidburhansModel sourceAnswered 100% · Latency 0.0s	5.1B	Open	38.5 points
64	Jev-Omniakhilaaa3Model sourceCalibration 16.1 · Answered 99.6% · Latency 0.1s	12B	Open	38.3 points
65	razorback16 openjev diffusiongemma (NVFP4, vLLM)Not listedCalibration 23.2 · Answered 99.6% · Latency 0.0s	26B	Evaluated configuration	37.5 points
66	Intelif Qwen3-4BNot listedAnswered 100% · Latency 0.0s	4.0B	Evaluated configuration	37.5 points
67	Jevstral 8BAmirBrahamModel sourceAnswered 100% · Latency 0.0s	~8.0B	Open	37.4 points
68	JobeNot listedCalibration 12.3 · Answered 100% · Latency 0.1s	4.7B	Evaluated configuration	37.3 points
69	Decider 2BMapikaModel sourceCalibration 7.7 · Answered 100% · Latency 0.0s	2.3B	Open	36.9 points
70	this-that 1.2Not listedCalibration 19.8 · Answered 99.6% · Latency 0.0s	1.9B	Evaluated configuration	36.4 points
71	Bespoke Nimble 9B v2bespokelabsModel sourceCalibration 2.4 · Answered 99.6% · Latency 0.1s	9.7B	Open	35.7 points
72	Gevva E4BdavidburhansModel sourceAnswered 100% · Latency 0.1s	8.0B	Open	34.9 points
73	vLLM-SR Decision 2.0 Eos 0.8BNot listedAnswered 100% · Latency 0.0s	870M	Evaluated configuration	34.6 points
74	EXAONE-4.0-1.2B-JEV v0.3Not listedAnswered 100% · Latency 0.1s	~1.3B	Evaluated configuration	34.0 points
75	Dinah-0LukitaduarteModel sourceAnswered 100% · Latency 0.0s	150M	Open	33.6 points
76	Kodiak v0.2 1B (accuracy)cortex-agent-llcModel sourceAnswered 99.9% · Latency 0.1s	3.0B	Open	31.7 points
77	Sifr 0.8B v3.1Not listedAnswered 100% · Latency 0.0s	870M	Evaluated configuration	31.6 points
78	OneJev 0.8BOmniJevModel sourceAnswered 100% · Latency 0.1s	870M	Open	30.9 points
79	Bosun v3.1 0.6BClause LogicModel sourceCalibration 14.2 · Answered 99.6% · Latency 0.0s	596M	Open	29.9 points
80	Bekko System One v0 68MhotchpotchModel sourceAnswered 100% · Latency 0.0s	~68M	Open	29.6 points
81	openvonsNot listedCalibration 37.1 · Answered 100% · Latency 0.0s	4.0B	Evaluated configuration	28.7 points
82	vLLM-SR Decision 2.0 Kai 0.6BNot listedAnswered 100% · Latency 0.0s	600M	Evaluated configuration	28.2 points
83	Sieve 2Bsthanika-aiModel sourceAnswered 100% · Latency 0.1s	2.3B	Open	28.1 points
84	Solomon v1.1Not listedCalibration 8.1 · Answered 75% · Latency 0.2s	28B	Evaluated configuration	26.6 points
85	open-jev (pngwn)Not listedCalibration 5.9 · Answered 99.8% · Latency 0.1s	4.7B	Evaluated configuration	25.4 points
86	Von 1.3Not listedAnswered 100% · Latency 0.0s	400M	Evaluated configuration	23.5 points
87	mmastrac diffusiongemma (vLLM PR 57250)Not listedCalibration 20.2 · Answered 86.8% · Latency 0.1s	26B	Evaluated configuration	23.4 points
88	Jev-Style 0.8B v3Not listedAnswered 99.9% · Latency 0.0s	870M	Evaluated configuration	23.0 points
89	JevK5 LitealibiserikbayModel sourceAnswered 69% · Latency 0.0s	300M	Open	22.7 points
90	GLiNER2.5-DecidefastinoModel sourceCalibration 8.8 · Answered 99.6% · Latency 0.0s	486M	Open	20.3 points
91	jeffLogan MarkewichCalibration 9.7 · Answered 100% · Latency 0.0s	~576M	Open	18.4 points
92	Kev 0.8B r15Not listedCalibration 4 · Answered 100% · Latency 0.0s	870M	Evaluated configuration	18.3 points
93	Tev1-4B-experimentaltogethercomputerModel sourceCalibration 10.4 · Answered 86.2% · Latency 0.0s	4.7B	Open	17.6 points
94	SemIfNot listedCalibration 9.9 · Answered 86.1% · Latency 0.1s	4.7B	Evaluated configuration	17.2 points
95	Intern-Decision-2BinternlmModel sourceCalibration 6.1 · Answered 99.6% · Latency 0.0s	2.3B	Open	17.1 points
96	Bekko System One v0 17MhotchpotchModel sourceAnswered 100% · Latency 0.0s	~17M	Open	16.4 points
97	LavoirmoganaiModel sourceCalibration 14.9 · Answered 99.7% · Latency 0.0s	396M	Open	15.6 points
98	JPT-0.8BkirpModel sourceCalibration 6.9 · Answered 100% · Latency 0.1s	873M	Open	15.1 points
99	Metask-Jev-4BWayfind (metask-ai)Model sourceCalibration 10.3 · Answered 86.8% · Latency 0.1s	4.7B	Open	13.8 points
100	DM-JEPA 1.1Not listedAnswered 91.5% · Latency 0.1s	149M	Evaluated configuration	12.8 points
101	system-one-gemmaNot listedCalibration 23.9 · Answered 99.9% · Latency 0.0s	268M	Evaluated configuration	11.2 points
102	mini-jevNot listedCalibration 34.6 · Answered 86.8% · Latency 0.1s	4.0B	Evaluated configuration	9.6 points
103	LayaConvai InnovationsModel sourceCalibration 14 · Answered 99.6% · Latency 0.0s	421M	Open	8.6 points
104	Intern-Decision-0.8BinternlmModel sourceCalibration 2.5 · Answered 99.6% · Latency 0.0s	873M	Open	7.9 points
105	CLM-v0.1-8BContrastive-LMModel sourceCalibration 32.3 · Answered 99.6% · Latency 0.0s	8.2B	Open	6.5 points
106	LFM2.5-2.6B-RLCDmonotykamaryModel sourceCalibration 25.4 · Answered 99.5% · Latency 0.0s	2.7B	Open	6.5 points
107	Julia 1SupersonicLabsModel sourceCalibration 42 · Answered 99.6% · Latency 0.0s	~141M	Open	5.3 points
108	Tev1-0.8B-experimentaltogethercomputerModel sourceCalibration 12.6 · Answered 86.2% · Latency 0.0s	873M	Open	3.8 points
109	Qwen-2.5-1B-RLCDharshathegModel sourceCalibration 20.5 · Answered 99.7% · Latency 0.0s	1.5B	Open	1.0 points
110	VerdictNot listedCalibration 15.4 · Answered 40.2% · Latency 0.0s	151M	Evaluated configuration	0.0 points
111	LFM2.5-350M-RLCDnotnotsamuelModel sourceCalibration 56.8 · Answered 100% · Latency 0.0s	354M	Open	0.0 points
112	ZevNot listedAnswered 100% · Latency 0.0s	Not disclosed	Evaluated configuration	0.0 points

How to read this leaderboard

The source uses a 0 to 100 scale, with 0 at chance and 100 at perfect performance. Scores are coverage-adjusted; an unanswered or unsupported request counts as wrong.

Operator receipt: 112 sourced rows are currently displayable on this page; the highest published score among these 112 evaluated systems is Perplexity Decider v1.1 (27B) at 64.7 points, which does not establish a market leader.

Honest limit: This is an area score from a composite that includes private tasks. It is display-only and does not enter general model rankings.

Decision Models

Decision Index 0.3 leaderboard Decision Index methodology

How we show Decision Index 0.3

The Decision Index 0.3 board lists 112 evaluated systems. This snapshot follows source commit f03e5a6 and keeps the retrieval score values and size labels published on the board.

The full score weights 37 public benchmarks at 20%, private tests of the same skills at 50%, and private tasks from new domains at 30%. The source reports 110,201 decisions. Scores are chance-corrected and coverage-adjusted; unsupported or unanswered requests count as wrong.

The published aggregates include private task results, but the task contents and item-level answers are not public. We show the board as external reference data, separate from general model rankings.

Snapshot

112 systems37 public benchmarks110,201 decisions20% public · 80% privateDisplay only

Decision Index 0.3 board Decision Index methodology Space README and model repositories Decision Models

The published Decision Index 0.3 Retrieval snapshot places Perplexity Decider v1.1 (27B) first at 64.7 points. The third row is 0.2 points behind. The broader top-10 range is 3.3 points, so many of the published results sit in a relatively narrow band.

112 evaluated systems are shown for Decision Index 0.3 Retrieval. The benchmark falls in the Decision Models category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. Decision Index 0.3 Retrieval is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About Decision Index 0.3 Retrieval

Year

2026

Tasks

Retrieval area of the Decision Index 0.3 evaluation

Format

Chance-corrected area score

Difficulty

Retrieval and classification decision tasks

The area score appears on the board beside the full score. The full score combines public benchmarks at 20%, private tests of the same skills at 50%, and private tasks from new domains at 30%. Private task contents and item-level results are not published.

Decision Index 0.3 leaderboard Public benchmark source

Freshness and provenance

Version

Decision Index 0.3

Refresh cadence

Pinned benchmark-owner snapshot

Staleness state

Current

Question availability

37 public benchmarks; same-skill and new-domain task contents and item-level results are private

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does Decision Index 0.3 Retrieval measure?

Decision Index 0.3 publishes chance-corrected retrieval scores for 112 evaluated systems.

Which model leads the published Decision Index 0.3 Retrieval snapshot?

Perplexity Decider v1.1 (27B) currently leads the published Decision Index 0.3 Retrieval snapshot with 64.7 points retrieval score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on Decision Index 0.3 Retrieval?

The October 6, 2026 snapshot contains 112 AI models.

Last updated: October 6, 2026 · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 5,600+ readers.

One email each week. Unsubscribe anytime.