Task ranking

Best Decision Model for Agent Routing and Scoring

H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim) is the provisional first pick for Best Decision Model for Agent Routing and Scoring as of November 15, 2026, while independent task measurements are still limited.[1][2] It scores 77.94 on JevBench (#1 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 60 and its calibration axis 90, each out of 100. Bobcat Flash 1.2 (Gemma-4-26B-A4B-it + two merged rank-64 LoRA adapters, typed-decision readout at the first answer position, FP8, self-hosted) is the next-ranked alternative. It scores 75.79 on JevBench (#2 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 66 and its calibration axis 81, each out of 100.

About this recommendation

Updated
Evidence through
Sources
8
Revision
v47

Comparing Jev alternatives or choosing a System One model? Read our guide to AI decision models, routing, classification and scoring for the decision model vs LLM tradeoffs and a self-hosting checklist.

Decision audit

Why this result

Inspect the inputs and the computed order behind the recommendation.

Models screened

24

live candidates

Evaluation feeds

1

task-weighted

Winner coverage

100%

intended feed weight

Largest provider share

1 of 8

chinhnc

Provisional source breadth. 1 citation families and 0 practitioner families provide attributed capability reports; 0 cautionary threads retained. The largest citation family contributes 100%.

H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim) leads the measured task comparison with limited independent task measurements. Community volume, release age and usage popularity add no ranking points.

Real-world track record is still developing. The recommendation can lead on independent measurements while practitioner reports accumulate.

Sources evaluated

These are the weights used in scoring. Missing measurements are omitted from each model’s average. A feed with no result for any model released in the last 60 days is stale and does not score until it catches up.

winner: H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim)
Evaluation feedWeightWinner resultField measured
JevBenchObserved 2026-11-15 · evaluation date not supplied · H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim)
100%
#124/24

Provider concentration

Each exact model is scored separately; provider identity is not a ranking input.

chinhnc13%
  • chinhnc1 model
  • h2oai1 model
  • juspay1 model
  • levanto-labs1 model
  • perplexity-ai1 model
  • sanghwa-na1 model
  • surogate1 model
  • torchcast-ai1 model

Decision table

Every published model is shown in computed order. Practitioner sources are distinct community threads, not the citations repeated in the prose below.

  1. h2oai · coverage 100% · provisional evidence

    #1 JevBench

  2. sanghwa-na · coverage 100% · provisional evidence

    #2 JevBench

  3. perplexity-ai · coverage 100% · provisional evidence

    #3 JevBench

  4. juspay · coverage 100% · provisional evidence

    #4 JevBench

  5. torchcast-ai · coverage 100% · provisional evidence

    #5 JevBench

  6. chinhnc · coverage 100% · provisional evidence

    #6 JevBench

  7. levanto-labs · coverage 100% · provisional evidence

    #7 JevBench

  8. surogate · coverage 100% · provisional evidence

    #8 JevBench

Task score averages the available, task-weighted measurements. Benchmark positions compare models matched to our catalog. Models measured by at least two independent evaluator families are considered first. Missing tests, practitioner report counts, popularity and release age add no points. Coverage describes evidence breadth, not a probability of correctness. Scores within two points are shown as close under this policy; this is not a statistical confidence interval. When the leading scores are that close, the model measured by more evaluator families ranks first.

  1. H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim) scores 77.94 on JevBench (#1 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 60 and its calibration axis 90, each out of 100.

    Best when: Its strongest result for this task is on JevBench, which measures calibrated typed decisions.
    Source 1
    “Within the 126 catalog models matched to this evaluation: Scores 77.94 on JevBench (#1 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 60 and its calibration axis 90, each out of 100.”

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  2. Bobcat Flash 1.2 (Gemma-4-26B-A4B-it + two merged rank-64 LoRA adapters, typed-decision readout at the first answer position, FP8, self-hosted) scores 75.79 on JevBench (#2 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 66 and its calibration axis 81, each out of 100.

    Best when: Its strongest result for this task is on JevBench, which measures calibrated typed decisions.
    Source 2
    “Within the 126 catalog models matched to this evaluation: Scores 75.79 on JevBench (#2 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 66 and its calibration axis 81, each out of 100.”

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  3. Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout) scores 74.93 on JevBench (#3 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 76 and its calibration axis 90, each out of 100.

    Best when: Its strongest result for this task is on JevBench, which measures calibrated typed decisions.
    Source 3
    “Within the 126 catalog models matched to this evaluation: Scores 74.93 on JevBench (#3 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 76 and its calibration axis 90, each out of 100.”

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  4. Xor 26B-A4B (Juspay, Gemma-4-26B-A4B, bf16) scores 74.01 on JevBench (#4 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 57 and its calibration axis 85, each out of 100.

    Best when: Its strongest result for this task is on JevBench, which measures calibrated typed decisions.
    Source 4
    “Within the 126 catalog models matched to this evaluation: Scores 74.01 on JevBench (#4 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 57 and its calibration axis 85, each out of 100.”

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  5. torchcast-decision-27b (Torchcast AI, Qwen3.8-27B fine-tune) scores 74 on JevBench (#5 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 70 and its calibration axis 90, each out of 100.

    Best when: Its strongest result for this task is on JevBench, which measures calibrated typed decisions.
    Source 5
    “Within the 126 catalog models matched to this evaluation: Scores 74 on JevBench (#5 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 70 and its calibration axis 90, each out of 100.”

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  6. Quyet-1.0-Large (Chinh Nguyen, Gemma-4-31B decoder, option-letter logits) scores 73.78 on JevBench (#6 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 73 and its calibration axis 90, each out of 100.

    Best when: Its strongest result for this task is on JevBench, which measures calibrated typed decisions.
    Source 6
    “Within the 126 catalog models matched to this evaluation: Scores 73.78 on JevBench (#6 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 73 and its calibration axis 90, each out of 100.”

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  7. Sage 1.3.0 (Levanto Labs) scores 73.73 on JevBench (#7 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 66 and its calibration axis 92, each out of 100.

    Best when: Its strongest result for this task is on JevBench, which measures calibrated typed decisions.
    Source 7
    “Within the 126 catalog models matched to this evaluation: Scores 73.73 on JevBench (#7 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 66 and its calibration axis 92, each out of 100.”

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

  8. Surogate Rune 26B-A4B v3 (v1.6 pool) scores 73.43 on JevBench (#8 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 56 and its calibration axis 91, each out of 100.

    Best when: Its strongest result for this task is on JevBench, which measures calibrated typed decisions.
    Source 8
    “Within the 126 catalog models matched to this evaluation: Scores 73.43 on JevBench (#8 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 56 and its calibration axis 91, each out of 100.”

    No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.

Frequently asked

What is the best model for Best Decision Model for Agent Routing and Scoring?
H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim) ranks first as of November 15, 2026. H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim) scores 77.94 on JevBench (#1 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 60 and its calibration axis 90, each out of 100.[1]
What is the best alternative to H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim)?
Bobcat Flash 1.2 (Gemma-4-26B-A4B-it + two merged rank-64 LoRA adapters, typed-decision readout at the first answer position, FP8, self-hosted) ranks second. Bobcat Flash 1.2 (Gemma-4-26B-A4B-it + two merged rank-64 LoRA adapters, typed-decision readout at the first answer position, FP8, self-hosted) scores 75.79 on JevBench (#2 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 66 and its calibration axis 81, each out of 100.[2]

Sources

  1. 1

    “Within the 126 catalog models matched to this evaluation: Scores 77.94 on JevBench (#1 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 60 and its calibration axis 90, each out of 100.”

    JevBench · Benchmark · Nov 15, 2026
  2. 2

    “Within the 126 catalog models matched to this evaluation: Scores 75.79 on JevBench (#2 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 66 and its calibration axis 81, each out of 100.”

    JevBench · Benchmark · Nov 15, 2026
  3. 3

    “Within the 126 catalog models matched to this evaluation: Scores 74.93 on JevBench (#3 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 76 and its calibration axis 90, each out of 100.”

    JevBench · Benchmark · Nov 15, 2026
  4. 4

    “Within the 126 catalog models matched to this evaluation: Scores 74.01 on JevBench (#4 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 57 and its calibration axis 85, each out of 100.”

    JevBench · Benchmark · Nov 15, 2026
  5. 5

    “Within the 126 catalog models matched to this evaluation: Scores 74 on JevBench (#5 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 70 and its calibration axis 90, each out of 100.”

    JevBench · Benchmark · Nov 15, 2026
  6. 6

    “Within the 126 catalog models matched to this evaluation: Scores 73.78 on JevBench (#6 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 73 and its calibration axis 90, each out of 100.”

    JevBench · Benchmark · Nov 15, 2026
  7. 7

    “Within the 126 catalog models matched to this evaluation: Scores 73.73 on JevBench (#7 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 66 and its calibration axis 92, each out of 100.”

    JevBench · Benchmark · Nov 15, 2026
  8. 8

    “Within the 126 catalog models matched to this evaluation: Scores 73.43 on JevBench (#8 of 126), Benchmark Heaven's composite of decision accuracy, calibration, speed, and cost for typed decision models; its decision accuracy axis is 56 and its calibration axis 91, each out of 100.”

    JevBench · Benchmark · Nov 15, 2026

Rankings synthesized from community evidence and open benchmarks. See methodology. Not driven by vendor marketing.