whatmodel.sh / decision / Perplexity Decider v1.1 (27B)

Perplexity Decider v1.1 (27B)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Perplexity Decider v1.1 (27B) is an open-weights decision model from Perplexity, released 5 October 2026. Licence: apache-2.0.

Catalogue identity notes

Explicit v1.1 release, kept separate from v1.2. Medical study records requested and returned v1.1 endpoint IDs. Public weight repository verified separately; equivalence between the hosted snapshot and a particular downloadable weight hash is not established.

Model information sources

Release date ↗ · Weights licence ↗

71.9±0.9#10 of 355 decision modelsfrom 73 benchmark results across 5 sources

$ how it compares

The decision models ranked either side of Perplexity Decider v1.1 (27B), out of 357 with a score, then more from Perplexity.

#modelscorewhereRAMreleased
8vLLM-SR d3-ultra (Decision 3.0, 27B) vLLM Semantic Router72±1model files
9vLLM-SR d3-turbo v1.1 (27B) vLLM Semantic Router72±1model files
10Perplexity Decider v1.1 (27B) Perplexity72±1api price unknown2026-10-05
11WeMM-JEV-27B Rover0071±1model files
12KnowLine 35B-A3B Gen1.5 PelaAI71±2model files2026-10-22
more from Perplexity
17Perplexity Decider v1.2 (27B) Perplexity70±1model files
90Perplexity Decider Flash v1.2 (4B) Perplexity62±1local~10GB

see all decision models · find the ones that fit your machine

$ what it's good at

retrieval & classification▁▁▁·▁▁▁▁▃▃█▅ #41 of 258 · 10% of its score · 9 tests

Trails Standard One 8B SH (v4), StartLux-Decision v1.1 2B, GEV-26B-Decide here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
DecisionPivot v1.1 · paws (accuracy)94.2% ✎█·██·······█ beats 3/3easy · avg 68%▇ 1.9%sharp
DecisionPivot v1.1 · nevir (pair_accuracy)80.2% ✎█··█··█····█ beats 3/3medium · avg 49%▇ 1.7%sharp
LeanMCP · nfcorpus rank noul38.1% ✎▃▃·····▃▃▅██ beats 10/11hard · avg 29%▇ 1.2%blunt · at its level
LeanMCP · scifact rank noul75.6% ✎▂····▂···▃▅█ beats 11/11medium · avg 62%▇ 1.1%typical
DecisionPivot v1.1 · entity match amazon google (auroc)0.9700 auroc ✎█···█·█····█ beats 3/3avg 0.290 auroc▇ 0.9%typical
DecisionPivot v1.1 · entity match abt buy (auroc)0.9970 auroc ✎█······█··██ beats 3/3avg 0.320 auroc▇ 0.9%typical
DecisionPivot v1.1 · rerank nfcorpus (ndcg@10)0.4410 ndcg@10 ✎█··········▄ beats 2/2avg -0.224 ndcg@10▇ 0.7%small run
DecisionPivot v1.1 · rerank scifact (ndcg@10)0.8140 ndcg@10 ✎█·········██ beats 1/2avg 0.162 ndcg@10▇ 0.7%small run
1 more test
testresultvs the others in its rundifficultycountswhy
GEO Citation · listwise Spearman (October 2026)0.5748 spearman-rho ✎▃···▃······█ beats 2/4avg 0.256 spearman-rho▇ 0.4%blunt
knowledge & reasoning▁▂▂▂▂▃▃▅▅▆█▃ #2 of 263 · 25% of its score · 15 tests

Beats Jeffan-31b-1710, baselabs mammoth-1 (27B), Sieve 31B v3 here, though they rank higher overall. Trails Jev 1.13 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
DecisionPivot v1.1 · primevul pair (pair_accuracy)6.6% ✎▄·········█▄ beats 3/3very hard · avg 1.2%▇ 2.4%sharp · too hard for it
Hallucination judging · Instruction inconsistency71.1% ✎▄····█·▄···▄ beats 4/4medium · avg 38%▇ 2.3%sharp · at its level
Hallucination judging · Context inconsistency82.7% ✎▄·▄···▄····█ beats 4/4medium · avg 49%▇ 2.2%sharp
LeanMCP · medqa usmle88.9% ✎▄▄▄·▄▄▄▄▄·██ beats 11/11easy · avg 72%▇ 1.9%typical
Hallucination judging · Logical inconsistency73.3% ✎█·█···█·█··█ beats 4/4medium · avg 43%▇ 1.9%sharp · at its level
LeanMCP · scienceqa text97.0% ✎▄··▄█▄█▄·█▄▄ beats 11/11very easy · avg 88%▇ 1.7%sharp · maxed out
DecisionPivot v1.1 · folio validation (accuracy)84.3% ✎█·········▄▄ beats 3/3medium · avg 56%▇ 1.7%sharp
LeanMCP · pubmedqa69.6% ✎▂···▂·▂·▄▄▂█ beats 8/11medium · avg 54%▇ 1.7%typical
7 more tests
testresultvs the others in its rundifficultycountswhy
LeanMCP · medmcqa75.7% ✎▃·▃··▃█▃·▅▅▃ beats 7/11medium · avg 63%▇ 1.6%typical
MedQA · Freyer et al. (first-valid accuracy)87.8% ✎█··▃▃·▃▃▃·▃█ beats 8/11easy · avg 72%▇ 1.5%typical
Hallucination judging · Factual contradiction78.9% ✎█···█··█··██ beats 4/4medium · avg 55%▇ 1.4%typical
DecisionPivot v1.1 · cuad clause (accuracy)82.8% ✎▄·········█▄ beats 3/3medium · avg 65%▇ 1.3%typical
LeanMCP · mmlu pro70.8% ✎▄··▄▄████··▄ beats 8/11medium · avg 54%▇ 1.2%typical
Hallucination judging · Factual fabrication79.8% ✎█·······▄▄·▄ beats 2/4medium · avg 59%▇ 1.1%typical
DecisionPivot v1.1 · legalbench mix (accuracy)78.5% ✎█····█···█·█ beats 2/3medium · avg 62%▇ 1.1%typical
workflows▁▁▁▂▁▂▂▆█▄▁▃ #1 of 114 · 65% of its score · 49 tests

Beats baselabs mammoth-1 (27B), StartLux-Decision v1.1 27B here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
DecisionPivot v1.1 · triage order (pair accuracy)33.8% ✎█···█······█ beats 2/2very hard · avg 6.4%▇ 2.8%sharp · at its level · small run
DecisionPivot v1.1 · filter match (pair accuracy)37.3% ✎█·····█····█ beats 2/2very hard · avg 4.9%▇ 2.7%sharp · at its level · small run
DecisionPivot v1.1 · sched conflict (pair accuracy)51.3% ✎█·····█····█ beats 2/2very hard · avg 7.1%▇ 2.6%sharp · at its level · small run
DecisionPivot v1.1 · mind2web element (accuracy)89.7% ✎█···█·█····█ beats 3/3medium · avg 54%▇ 2.4%sharp
DecisionPivot v1.1 · state goal (pair accuracy)76.7% ✎█···█······█ beats 2/2very hard · avg 11%▇ 2.3%sharp · at its level · noisy · small run
DecisionPivot v1.1 · record rule (pair accuracy)50.0% ✎█········█·█ beats 2/2very hard · avg 6.0%▇ 2.3%sharp · at its level · noisy · small run
DecisionPivot v1.1 · processbench (accuracy)75.3% ✎█·····█··█·█ beats 3/3medium · avg 41%▇ 2.2%sharp
DecisionPivot v1.1 · perm inherit (pair accuracy)78.7% ✎██·········█ beats 2/2very hard · avg 15%▇ 2.1%sharp · at its level · small run
41 more tests
testresultvs the others in its rundifficultycountswhy
DecisionPivot v1.1 · ledger reconcile (pair accuracy)41.3% ✎█·······█··█ beats 2/2very hard · avg 12%▇ 2.1%sharp · at its level · small run
DecisionPivot v1.1 · judgebench (accuracy)82.4% ✎█······██··█ beats 3/3medium · avg 52%▇ 2.1%sharp
DecisionPivot v1.1 · log count check (pair accuracy)72.0% ✎█·····█····█ beats 2/2very hard · avg 4.6%▇ 2.0%sharp · at its level · noisy · small run
DecisionPivot v1.1 · llmbar (accuracy)88.2% ✎██·····█···█ beats 3/3medium · avg 58%▇ 1.9%sharp
DecisionPivot v1.1 · answerable facts (pair accuracy)92.0% ✎█·█········█ beats 2/2hard · avg 18%▇ 1.9%sharp · at its level · noisy · small run
DecisionPivot v1.1 · rewardbench2 (accuracy)81.9% ✎█·····█···██ beats 3/3medium · avg 57%▇ 1.8%sharp
LeanMCP · mcpmark server routing58.0% ✎▂··▂··▄█▂·▂▄ beats 11/11medium · avg 36%▇ 1.7%at its level
DecisionPivot v1.1 · clause change (pair accuracy)96.0% ✎█·█········█ beats 2/2very hard · avg 10.0%▇ 1.7%sharp · at its level · noisy · small run
DecisionPivot v1.1 · toolcall valid (pair accuracy)90.0% ✎█··········▄ beats 2/2hard · avg 27%▇ 1.7%sharp · small run
DecisionPivot v1.1 · browsesafe (accuracy)80.5% ✎▄·······▄··█ beats 3/3medium · avg 56%▇ 1.7%sharp
DecisionPivot v1.1 · squad2 answerable (accuracy)91.8% ✎███········█ beats 3/3easy · avg 69%▇ 1.6%sharp
DecisionPivot v1.1 · dep resolve (pair accuracy)94.7% ✎██·········█ beats 2/2very hard · avg 6.8%▇ 1.6%sharp · at its level · noisy · small run
DecisionPivot v1.1 · policy case (pair accuracy)97.5% ✎██·········█ beats 2/2very hard · avg 14%▇ 1.5%sharp · at its level · noisy · small run
DecisionPivot v1.1 · rewardbench (accuracy)95.0% ✎█······█·█·█ beats 2/3easy · avg 83%▇ 1.4%sharp · maxed out
LeanMCP · mcpbench server routing98.2% ✎█·███··▄···▄ beats 8/9easy · avg 82%▇ 1.3%sharp · maxed out
LeanMCP · rewardbench281.5% ✎▄▄▄·▄▄▄▄█·█▄ beats 11/11medium · avg 58%▇ 1.3%typical
DecisionPivot v1.1 · cap boundary record (MSE skill)0.1750 skill ✎█··█····█··█ beats 2/3avg -0.616 skill▇ 1.2%typical
LeanMCP · aegis2 response80.8% ✎█··▃·▃·█··▅▅ beats 8/11easy · avg 70%▇ 1.1%typical
DecisionPivot v1.1 · cap boundary card (MSE skill)0.0940 skill ✎█··█····█··█ beats 3/3avg -0.709 skill▇ 1.1%typical
LeanMCP · helpsteer2 helpfulness0.8689 ordinal levels ✎▄▄▄▄▄·▄·█▄▄▄ beats 10/10avg 1.357 ordinal levels▇ 1.1%blunt
DecisionPivot v1.1 · freshqa premise (auroc)0.9560 auroc ✎█·······█·██ beats 3/3avg 0.209 auroc▇ 1.0%typical
DecisionPivot v1.1 · faithbench (accuracy)55.6% ✎█·█·█······█ beats 2/3medium · avg 38%▇ 1.0%at its level
LeanMCP · sst50.5078 ordinal levels ✎▂·····▂▂▄█▄▂ beats 9/11avg 0.879 ordinal levels▇ 1.0%blunt
DecisionPivot v1.1 · contextual judge (accuracy)59.6% ✎█·██·······█ beats 2/3medium · avg 43%▇ 0.9%at its level
DecisionPivot v1.1 · agentreward success (accuracy)78.1% ✎█···██·····█ beats 3/3medium · avg 58%▇ 0.9%typical
DecisionPivot v1.1 · jbb judge (auroc)0.9740 auroc ✎█····█··█··█ beats 2/3avg 0.299 auroc▇ 0.9%typical
DecisionPivot v1.1 · xstest (auroc)0.9960 auroc ✎█········▄·▄ beats 3/3avg 0.320 auroc▇ 0.9%typical
DecisionPivot v1.1 · known unknown (accuracy)86.3% ✎█··█······██ beats 3/3easy · avg 69%▇ 0.9%typical
DecisionPivot v1.1 · aegis2 prompt (accuracy)85.2% ✎█·····█·█··█ beats 3/3easy · avg 73%▇ 0.8%typical
DecisionPivot v1.1 · label errors (accuracy)60.9% ✎██·······█·█ beats 1/3medium · avg 47%▇ 0.7%at its level
DecisionPivot v1.1 · bipia injection (accuracy)92.0% ✎█····█····██ beats 1/3easy · avg 76%▇ 0.7%sharp · maxed out · noisy
DecisionPivot v1.1 · cosapien policy (accuracy)87.8% ✎▄·······█··▄ beats 2/3easy · avg 75%▇ 0.7%typical
LeanMCP · atbench50074.6% ✎▃··▅▅█··▅··▅ beats 4/11medium · avg 63%▇ 0.7%sharp · noisy
LeanMCP · prompt injections90.5% ✎▂▂▂▂█·▄▂···▃ beats 11/11medium · avg 62%▇ 0.6%typical
DecisionPivot v1.1 · jailbreak wild (accuracy)74.3% ✎█······█·█·█ beats 0/3easy · avg 65%▇ 0.6%typical
DecisionPivot v1.1 · bfcl relevance (accuracy)88.6% ✎█··█······██ beats 1/3easy · avg 82%▇ 0.5%maxed out
DecisionPivot v1.1 · ambigqa (accuracy)52.0% ✎█·······▄··▄ beats 0/3medium · avg 44%▇ 0.5%at its level
DecisionPivot v1.1 · selfaware (accuracy)59.7% ✎█····▄·····▄ beats 1/3medium · avg 53%▇ 0.3%blunt · at its level
LeanMCP · mcpbench tool selection100.0% ✎▂··········█ beats 6/9very easy · avg 89%▇ 0.3%sharp · maxed out · noisy
LeanMCP · mcpmark tool selection100.0% ✎▂··········█ beats 6/11very easy · avg 96%▇ 0.1%sharp · maxed out · noisy
LeanMCP · jailbreak classification89.8% ✎▃·▄··▂···█▆▂ beats 0/11very easy · avg 95%▇ 0.0%blunt · maxed out · noisy

$ where to run it

Through an API

api · OpenRouter · price unknown · openrouter.ai ↗

Download model files ↗
Public, ungated model files verified. Named release; historical hosted weights are not proven byte-identical. Custom native decision head, so no unverified chat-model run command supplied. Source · checked 2026-10-31.

Model repository and documentation: huggingface.co/perplexity-ai ↗

$ how the score was worked out · DecisionPivot · StartLux Labs 69%, LeanMCP · JevBench 20%, Decision Models as Judges for… 9% and 2 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
DecisionPivot · StartLux Labs48CC BY 4.0
LeanMCP · JevBench18MIT
Decision Models as Judges for Hallucination Detection5CC-BY-4.0
Freyer et al. · MedQA1CC BY 4.0
Citead GEO Citation Benchmark (October 2026)1CC BY-SA 4.0
prior · pull toward the middle
weighted average= 71.9

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json