Jared PalmerDecision modelOpen source

Kev 9B

PopularityRank 274 of 876

40out of 100-2 vs. last month

Summary

Kev 9B is the largest current Kev adapter variant. Its LoRA adapter and pointer head sit on a 9-billion-parameter base model and return calibrated Choice, Noul, and Score probabilities for typed questions. Like the smaller variants, it makes the decision in one pass.

The family overview names L40S and H100 as target hardware. The model card validates Kev 9B up to 8,192 tokens, while the server accepts states up to 65,536 tokens; that does not establish unchanged model accuracy. The smaller adapter variants are Kev 4B and Kev 0.8B.

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 95 published benchmark results

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

JEVal (arXiv:2610.03935v2), overall choice/noul evaluation (Show measurement, test conditions, and source)

Source value: 70.87
Score: 70.87
Metric: accuracy (invalid outputs count as incorrect)
Unit: %
Benchmark version: arXiv:2610.03935v2, Table 1
Category: JEVal overall choice/noul
Direction: Higher is better
Sample: 10938
Source type: community-reported
Evaluator: Duan et al., General Decision Models: Benchmarking and Insights Beyond Jev (arXiv:2610.03935v2)
Status: active
Retrieved at: 2026-10-18
Methodology: Community-reported result. JEVal Table 1, exact reported model label Kev-9b; decision-model run, no reasoning-effort setting reported; local vLLM on NVIDIA H100; context window 32,768 tokens; input budget 24,576 tokens; temperature 2.2974; 10,938 choice and noul questions.
Context: No immutable evaluated-weight or serving-revision digest is given. Table 1 counts invalid outputs as incorrect for accuracy; ECE and raw Brier are computed only on valid responses, whose per-model N is not reported. The paper reports no confidence interval for these point values.

Source Duan et al., General Decision Models: Benchmarking and Insights Beyond Jev (arXiv:2610.03935v2)

StreamDecisionBench (Show measurement, test conditions, and source)

Source value: 28.76
Score: 28.76
Metric: Mean log-AUC 0.5–8 s
Unit: %
Direction: No direction published
Sample: 480
Source type: independent-evaluator
Evaluator: StreamDecisionBench GitHub README
Status: active
Retrieved at: 2026-10-18
Methodology: Independent evaluation. Self-hosted open-weight setting, exact source row “Kev-9B”. Three passes over all 480 states per pass (8 scenarios in 4 families), sampled at a 2 s step; the source reports means. The README says all 17 self-hosted settings use BF16 backbones, each on one RTX PRO 6000 (96 GB), with benchmark and model on the same GPU host. Same-host latency includes request processing, runtime queueing, and inference, and excludes download, initialization, and warmup. Serving stack, exact checkpoint revision, and other runtime settings are not stated.

Source StreamDecisionBench GitHub README

StreamDecisionBench (Show measurement, test conditions, and source)

Source value: 29.86
Score: 29.86
Metric: Mean untimed (%)
Unit: %
Direction: No direction published
Sample: 480
Source type: independent-evaluator
Evaluator: StreamDecisionBench GitHub README
Status: active
Retrieved at: 2026-10-18
Methodology: Independent evaluation. Self-hosted open-weight setting, exact source row “Kev-9B”. Three passes over all 480 states per pass (8 scenarios in 4 families), sampled at a 2 s step; the source reports means. The README says all 17 self-hosted settings use BF16 backbones, each on one RTX PRO 6000 (96 GB), with benchmark and model on the same GPU host. Same-host latency includes request processing, runtime queueing, and inference, and excludes download, initialization, and warmup. Serving stack, exact checkpoint revision, and other runtime settings are not stated.

Source StreamDecisionBench GitHub README

Knowledge and evidence exams (Show measurement, test conditions, and source)

Source value: 78
Score: 78
Metric: accuracy
Unit: %
Benchmark version: Introducing Kev, published 2026-10-01; exact weight revision not stated in article
Direction: No direction published
Sample: 1046
Source type: vendor-reported
Evaluator: Jared Palmer, Introducing Kev
Status: active
Retrieved at: 2026-10-18
Methodology: Vendor-reported result. Development split; row covers buried facts and missing evidence. Author's evaluation harness, not an independent leaderboard. The article provides no full prompt, hardware, or inference-setting record for these cells.

Source Jared Palmer, Introducing Kev

JEVal aida-testc accuracy (Show measurement, test conditions, and source)

Source value: 96
Score: 96
Metric: accuracy (ACC)
Unit: %
Benchmark version: arXiv:2610.03935v2
Category: decision
Direction: Higher is better
Sample: 200
Source type: independent-evaluator
Evaluator: Duan et al., General Decision Models: Benchmarking and Insights Beyond Jev
Status: active
Retrieved at: 2026-10-22
Methodology: Independent evaluation. JEVal Table 10, general knowledge and reasoning, aida-testc (N = 200). The table identifies the model label and size; it does not state an immutable weight or serving revision.
Context: This is a dataset-specific component of JEVal, not the Table 1 aggregate. An exact weight revision and immutable per-model serving identity are absent. No confidence interval is reported in this row.

Source Duan et al., General Decision Models: Benchmarking and Insights Beyond Jev

JevAny external matrix: MMLU-Pro (ekzhang external panel, 1,000 questions) (Show measurement, test conditions, and source)

Source value: 50.7
Score: 50.7
Metric: all-requested accuracy; rejected or missing knowable questions count as wrong
Unit: %
Benchmark version: JevAny external decision-suite matrix, 2026-09-27; Kev repository revision 5920c5fe4ca8e0970ed4209ac2c9b8e18bea5109; checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c; panel MMLU-Pro (ekzhang external panel, 1,000 questions)
Category: decision
Direction: Higher is better
Sample: 1000
Source type: independent-evaluator
Evaluator: JevAny external decision-suite evaluation
Status: active
Retrieved at: 2026-10-22
Methodology: Independent evaluation. The published Kev checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c was evaluated on the external MMLU-Pro (ekzhang external panel, 1,000 questions) panel in JevAny's September 27, 2026 matrix. The report states that no sampling was performed.

Source JevAny external decision-suite evaluation

JevAny external matrix: WANLI-v2 (Show measurement, test conditions, and source)

Source value: 73.95
Score: 73.95
Metric: accuracy
Unit: %
Benchmark version: JevAny external decision-suite matrix, 2026-09-27; Kev repository revision 5920c5fe4ca8e0970ed4209ac2c9b8e18bea5109; checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c; panel WANLI-v2
Category: decision
Direction: Higher is better
Source type: independent-evaluator
Evaluator: JevAny external decision-suite evaluation
Status: active
Retrieved at: 2026-10-22
Methodology: Independent evaluation. The published Kev checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c was evaluated on the external WANLI-v2 panel in JevAny's September 27, 2026 matrix. The report states that no sampling was performed.

Source JevAny external decision-suite evaluation

JevAny external matrix: SemIf (Show measurement, test conditions, and source)

Source value: 90.97
Score: 90.97
Metric: accuracy
Unit: %
Benchmark version: JevAny external decision-suite matrix, 2026-09-27; Kev repository revision 5920c5fe4ca8e0970ed4209ac2c9b8e18bea5109; checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c; panel SemIf
Category: decision
Direction: Higher is better
Source type: independent-evaluator
Evaluator: JevAny external decision-suite evaluation
Status: active
Retrieved at: 2026-10-22
Methodology: Independent evaluation. The published Kev checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c was evaluated on the external SemIf panel in JevAny's September 27, 2026 matrix. The report states that no sampling was performed.

Source JevAny external decision-suite evaluation

JevAny external matrix: scienthoon (Show measurement, test conditions, and source)

Source value: 75.49
Score: 75.49
Metric: accuracy
Unit: %
Benchmark version: JevAny external decision-suite matrix, 2026-09-27; Kev repository revision 5920c5fe4ca8e0970ed4209ac2c9b8e18bea5109; checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c; panel scienthoon
Category: decision
Direction: Higher is better
Source type: independent-evaluator
Evaluator: JevAny external decision-suite evaluation
Status: active
Retrieved at: 2026-10-22
Methodology: Independent evaluation. The published Kev checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c was evaluated on the external scienthoon panel in JevAny's September 27, 2026 matrix. The report states that no sampling was performed.

Source JevAny external decision-suite evaluation

JevAny external matrix: TypeSafe (Show measurement, test conditions, and source)

Source value: 72.79
Score: 72.79
Metric: equal-case agreement
Unit: %
Benchmark version: JevAny external decision-suite matrix, 2026-09-27; Kev repository revision 5920c5fe4ca8e0970ed4209ac2c9b8e18bea5109; checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c; panel TypeSafe
Category: decision
Direction: Higher is better
Source type: independent-evaluator
Evaluator: JevAny external decision-suite evaluation
Status: active
Retrieved at: 2026-10-22
Methodology: Independent evaluation. The published Kev checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c was evaluated on the external TypeSafe panel in JevAny's September 27, 2026 matrix. The report states that no sampling was performed.

Source JevAny external decision-suite evaluation

JevBench public easy tier accuracy (Show measurement, test conditions, and source)

Source value: 100
Score: 100
Metric: accuracy
Unit: %
Benchmark version: JevAny external decision-suite matrix, 2026-09-27; JevBench public panel jevbench/easy; JevBench revision d06ee95988da1350eb8ae5511daa0e06bfffe911; checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c
Category: decision
Direction: Higher is better
Sample: 48
Source type: independent-evaluator
Evaluator: JevAny external decision-suite evaluation
Status: active
Retrieved at: 2026-10-22
Methodology: Independent evaluation. The published Kev checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c was evaluated in JevAny's September 27, 2026 matrix on the public JevBench easy tier, N=48.

Source JevAny external decision-suite evaluation

JevBench public original tier accuracy (Show measurement, test conditions, and source)

Source value: 90.28
Score: 90.28
Metric: accuracy
Unit: %
Benchmark version: JevAny external decision-suite matrix, 2026-09-27; JevBench public panel jevbench/original; JevBench revision d06ee95988da1350eb8ae5511daa0e06bfffe911; checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c
Category: decision
Direction: Higher is better
Sample: 72
Source type: independent-evaluator
Evaluator: JevAny external decision-suite evaluation
Status: active
Retrieved at: 2026-10-22
Methodology: Independent evaluation. The published Kev checkpoint jaredpalmer/kev-9b@2629c06a5aeb0feb3b9783bafed17ed8f39ecf5c was evaluated in JevAny's September 27, 2026 matrix on the public JevBench original tier, N=72.

Source JevAny external decision-suite evaluation

Page 1 of 8

Popularity compared

How often the model is searched for, used and mentioned, compared with every model of its category.

Popularity index

0 to 100, higher is more popular

20 of 37 models

AutoTrust JEV-27B-VL70

Clef52

GLiNER2.5-multi-Decide45

Decider 4B43

Jev43

pplx-decider-v1.1-27b42

Intern-Decision-4B40

Laya Typed Decisions40

Kev 9B40

pplx-decider-v1-27b39

Decider 0.8B39

Julia 139

Intern-Decision-0.8B38

GLiNER2.5-multi-Decide

Decider 4B

Jev

pplx-decider-v1.1-27b

Kev 9B

As of November 2026 · Source: gradually.ai popularity index

Over time

Weekly values from October 18, 2026 to November 6, 2026

Kev 9BIntern-Decision-4B 40Laya Typed Decisions 40pplx-decider-v1-27b 39Decider 0.8B 39

Method changes in the timeline, a jump at these points is no market event:

1November 1, 2026: Value only with search or usage data, 30-day attention

What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.

Kev 9B compared

Show all 15 comparisons

Similar models

More models of the same kind as Kev 9B, chosen by model line, developer, size, price and release date.

All specifications

Architecture

Model class: Typed-decision LoRA adapter with pointer readout; Source: Model class
Parameters: 9 billion base model; Source: Parameters
Base model: Qwen/Qwen3.5-9B-Base; Source: Base model

Input and output

Input: Text or structured states and typed questions; Source: Input
Output: Structured probabilities, no text generation; Source: Output

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking: Unknown; no source
Watermarking technique: Not known; no source

Licence and access

Access: Local; Source: Access
License: Apache-2.0 for the adapter and head; Source: License

API

Model ID: jaredpalmer/kev-9b; Source: Model ID

Further details

Result: Choice, Noul and Score with calibrated probabilities; Source: Result
Scope: Local Kev family models with a TypeSafe-compatible interface.; Source: Scope
Notes: The weights are adapters on top of the respective Qwen base models.; Source: Notes
Adapter: LoRA, rank 16; Source: Adapter

Sources and data date

Show all 22 sources

Benchmark sources at a glance

Which independent evaluators tested this model. This is coverage, not a quality ranking.

No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.

Without vendor claims and historical or disputed results; those stay labelled in the single values.

Type	Evidence and data date
Research status	Research date October 15, 2026. 13 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Unresolved	score-run configuration Original wording The source identifies the exact model row and some adjacent model/route settings, but does not bind all per-cell prompt, decoding, or serving details to the score table.
Unresolved	parenthetical ranges Original wording Ranges are printed beside each score but the article does not define them as confidence intervals or identify another statistical meaning.
Unresolved	Task done check sample count Original wording The source reports 155 of 204 items for this task, but the denominator meaning is not made clear; no sampleCount is asserted.
Unresolved	exact-checkpoint-and-inference-settings Original wording The article identifies model labels and the Kev 1.0 release context but gives no exact weight revision or full prompt/hardware/inference settings for these panel scores.
Not published	per-dataset-calibration-sample-count Original wording For the per-dataset averaged calibration-error aggregate, the article does not state a metric-specific evaluated N.
Not published	Immutable Decision Index served checkpoint revision Note The current public index gives a model URL and, where available, a served_checkpoint label, but no immutable commit/digest for the exact evaluated weights.
Additional source	Kev repository (retrieved September 24, 2026) · Kev repository · Editorial description reviewed October 2, 2026
Additional source	Kev release plan (retrieved September 24, 2026) · Kev release plan · Editorial description reviewed October 2, 2026
Additional source	Kev 9B checkpoint (retrieved September 24, 2026) · Kev 9B checkpoint · Editorial description reviewed October 2, 2026
Additional source	Kev 9b model card (retrieved October 15, 2026; October 21, 2026; October 23, 2026)
Additional source	Jared Palmer, Kev model card (retrieved October 16, 2026)
Additional source	Jared Palmer, Kev-9B v2 model card (retrieved October 18, 2026)
Additional source	Duan et al., General Decision Models: Benchmarking and Insights Beyond Jev (arXiv:2610.03935v2) (retrieved October 18, 2026)
Additional source	AIMultiple decision-models evaluation (retrieved October 18, 2026)
Additional source	StreamDecisionBench GitHub README (retrieved October 18, 2026)
Additional source	Jared Palmer, Introducing Kev (retrieved October 18, 2026)
Additional source	Duan et al., General Decision Models: Benchmarking and Insights Beyond Jev (retrieved October 22, 2026)
Additional source	JevAny external decision-suite evaluation (retrieved October 22, 2026)
Additional source	DecideBench (retrieved October 22, 2026)
Additional source	Kevala decision benchmark report (retrieved October 23, 2026)
Additional source	akitaonrails/jev-docs-benchmarks, bilingual v2 results (retrieved October 23, 2026)
Additional source	Jev Decision Index public index.json (release-v4) (retrieved October 28, 2026)
Additional source	OpenRouter: Popularity index (derived), CC BY 4.0
Additional source	Vercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional source	Ollama: cumulative downloads for the popularity index
Additional source	Replicate: cumulative runs for the popularity index
Additional source	Google Trending Now: limited feed selection for Germany and the US
Additional source	Creative Commons Attribution 4.0