d1 (hosted)

+ contribute

d1 (hosted) is a closed-weights decision model from Liquid AI. Licence: proprietary.

Catalogue identity notes

System One Bench identifies the tested endpoint as liquid/d1. Hosted service snapshot, not an immutable weight revision. Separate from chat or open-weight models with similar names. API access verified 2026-10-31; benchmark dates are retained per result. OpenRouter announces retirement on 2026-11-20; still available at review time. Not d1-3B or d1-omni-600M.

64.5±1.1#61 of 355 decision modelsfrom 49 benchmark results across 7 sources

$ how it compares

The decision models ranked either side of d1 (hosted), out of 357 with a score, then more from Liquid AI.

#	model	score	where	released
59	jev-brain · Gemma-4-31B (decision readout) wcole3	65±1	source code
60	Aleria Judge 1.1.0 (Gemma 4 31B) (decision readout) Decision Index submission: aleria-judge-1.1.0	65±4	source code	2026-10-14
61	d1 (hosted) Liquid AI	65±1	api price unknown
62	KnowLine 4B Gen5.5 PelaAI	64±1	model files	2026-10-22
63	Jev 1.14 TypeSafe	64±1	api $0.042/M input tokens	2026-10-14
	more from Liquid AI
21	d1-ultra Liquid AI	69±2	api price unknown	2026-10-15
32	d1.1 (hosted) Liquid AI	68±2	api price unknown
198	LiquidAI d1-3B Liquid AI	51±1	model files

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

$ what it's good at

retrieval & classification▁▁▁·▁▁▁▁▃▃█▅ #38 of 258 · 11% of its score · 8 tests

Beats Perplexity Decider v1.1 (27B), H2O Lightning 31B v1.1, AutoTrust-JEV-27B v1.1 here, though they rank higher overall. Trails StartLux-Decision v1.1 2B, GEV-26B-Decide, JEV-35B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
System One Bench: MIRACL pairwise (Arabic)	91.0% ✎	▂···▁··▃▄▂█▇ beats 26/29	easy · avg 84%	▇ 2.6%	typical
System One Bench: MASSIVE intent (Arabic)	90.0% ✎	▄·▁·▁▃▃▃█▄▇▇ beats 25/29	easy · avg 76%	▇ 2.3%	typical
LeanMCP · scifact rank noul	73.0% ✎	▂····▂···▃▅█ beats 6/11	medium · avg 62%	▇ 2.0%	at its level
LeanMCP · nfcorpus rank noul	38.3% ✎	▃▃·····▃▃▅██ beats 11/11	hard · avg 29%	▇ 1.8%	blunt · at its level
System One Bench: MIRACL rerank (Arabic)	90.5% ✎	▁▂·▁▁·▂·▄▄█▇ beats 29/29	easy · avg 79%	▇ 1.5%	typical
System One Bench: MASSIVE scenario (Arabic)	73.0% ✎	▃▁·▄▃▂█▂▁▁·▃ beats 21/29	medium · avg 60%	▇ 1.0%	at its level
System One Bench: AG News (English)	90.0% ✎	▁···▅▁▅▂▆█▂▅ beats 17/29	easy · avg 88%	▇ 0.3%	blunt
System One Bench: Banking77 (English)	89.0% ✎	█▅▆█▅▃▃▃···▂ beats 15/26	very easy · avg 87%	▇ 0.0%	blunt · noisy

language understanding▂▂▂▃▂▅▆▃█▆▄▃ #32 of 256 · 7.0% of its score · 9 tests

Beats KnowLine 31B Gen1, Ainvest-Decision-27B v1.0, H2O Lightning 31B v1.1 here, though they rank higher overall. Trails Clef-flash, Clef-omni, vLLM-SR d3-flash (9B) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
System One Bench: BoolQ (English)	90.0% ✎	▁▁▁▃·▁▃▃▇▇█▅ beats 16/29	easy · avg 86%	▇ 1.8%	typical
System One Bench: OSACT4 offensive (Arabic)	95.5% ✎	▁······▁▃█▇▄ beats 28/29	easy · avg 84%	▇ 1.6%	sharp
System One Bench: SST-2 (English)	95.5% ✎	▁·▃▁▂▂▃▅█·▂▁ beats 21/29	easy · avg 92%	▇ 1.1%	typical
System One Bench: SST-5 (English)	56.5% ✎	▃·▁▁▃·▇▄▄▄█▅ beats 17/29	medium · avg 51%	▇ 1.0%	blunt · at its level
System One Bench: XNLI Arabic (Arabic)	80.0% ✎	▄▁▃▁▄▇█▅▃▁·▃ beats 25/29	medium · avg 68%	▇ 0.5%	at its level
System One Bench: XNLI French (French)	84.7% ✎	▂▅·▁▁▅▅█▃▁·▃ beats 23/29	medium · avg 76%	▇ 0.5%	blunt · at its level
System One Bench: Tweet sentiment (Arabic)	67.5% ✎	▁······▁▁▄█▄ beats 15/29	medium · avg 62%	▇ 0.4%	blunt · at its level
System One Bench: XNLI English (English)	91.3% ✎	▃·▁·▃▇█▄█▃▃▁ beats 22/29	easy · avg 85%	▇ 0.1%	blunt · noisy

1 more test

test	result	vs the others in its run	difficulty	counts	why
System One Bench: DAIR Emotion (English)	63.5% ✎	██▃·····▁··▁ beats 25/29	medium · avg 59%	▇ 0.0%	blunt · at its level

knowledge & reasoning▁▂▂▂▂▃▃▅▅▆█▃ #12 of 263 · 50% of its score · 13 tests

Beats Ainvest-Decision-27B v1.0, StartLux-Decision v1.1 27B, vLLM-SR d3-turbo v1.1 (27B) here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
Etsabary · 1342 text decisions	71.9% ✎	▁··▂·▃▅▃▃▂▂█ beats 10/24	medium · avg 47%	▇▇▇▇▇ 12%	sharp · at its level
Hallucination judging · Context inconsistency	75.8% ✎	▄·▄···▄····█ beats 2/4	medium · avg 49%	▇▇ 4.3%	sharp · at its level
Hallucination judging · Instruction inconsistency	62.5% ✎	▄····█·▄···▄ beats 2/4	medium · avg 38%	▇▇ 4.1%	sharp · at its level
LeanMCP · scienceqa text	95.6% ✎	▄··▄█▄█▄·█▄▄ beats 8/11	very easy · avg 88%	▇▇ 4.1%	sharp · maxed out
LeanMCP · medqa usmle	88.1% ✎	▄▄▄·▄▄▄▄▄·██ beats 10/11	easy · avg 72%	▇▇ 3.9%	typical
Hallucination judging · Logical inconsistency	58.5% ✎	█·█···█·█··█ beats 0/4	medium · avg 43%	▇ 3.3%	sharp · at its level
CONFLICTS · source conflict types	62.6% ✎	▃▃··▅▃···▃▃█ beats 4/9	medium · avg 45%	▇ 3.2%	at its level
MedQA · Freyer et al. (first-valid accuracy)	88.4% ✎	█··▃▃·▃▃▃·▃█ beats 10/11	easy · avg 72%	▇ 3.0%	typical

5 more tests

test	result	vs the others in its run	difficulty	counts	why
LeanMCP · pubmedqa	70.6% ✎	▂···▂·▂·▄▄▂█ beats 10/11	medium · avg 54%	▇ 2.9%	at its level
LeanMCP · medmcqa	76.3% ✎	▃·▃··▃█▃·▅▅▃ beats 10/11	medium · avg 63%	▇ 2.8%	typical
Hallucination judging · Factual contradiction	67.8% ✎	█···█··█··██ beats 0/4	medium · avg 55%	▇ 2.6%	at its level
LeanMCP · mmlu pro	69.5% ✎	▄··▄▄████··▄ beats 7/11	medium · avg 54%	▇ 2.2%	at its level
Hallucination judging · Factual fabrication	70.5% ✎	█·······▄▄·▄ beats 0/4	medium · avg 59%	▇ 2.0%	typical

vision▁·▃▂▃▃▇▄▄█▄▃ #13 of 79 * · 8.3% of its score · 1 test

Beats Fastino GLiDE 1.2, StartLux-Decision v1.1 27B, vLLM-SR d3-turbo v1.1 (27B) here, though they rank higher overall. Trails Mercury Decide (free endpoint), Clef, Jev 1.14 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Etsabary · 200 image decisions	67.5% ✎	▂▂··▂▃▃·█▃▂▂ beats 5/14	medium · avg 45%	▇▇▇ 8.3%	sharp · at its level

workflows▁▁▁▂▁▂▂▆█▄▁▁ #18 of 114 · 22% of its score · 18 tests

Beats Nace.AI Drex Max v1 (27B), baselabs elephant-1 (9B), d1.1 (hosted) here, though they rank higher overall. Trails GPT-6 Luna Decisions, reflex-27b (DecisionBench 1d4bf0f2), Nace.AI Drex v2.0 (35B-A3B) here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LeanMCP · mcpbench server routing	92.0% ✎	█·███··▄···▄ beats 2/9	easy · avg 82%	▇▇ 3.8%	sharp · maxed out
LeanMCP · mcpmark server routing	45.9% ✎	▂··▂··▄█▂·▂▄ beats 2/11	medium · avg 36%	▇ 2.8%	at its level
LeanMCP · rewardbench2	73.6% ✎	▄▄▄·▄▄▄▄█·█▄ beats 6/11	medium · avg 58%	▇ 2.4%	at its level
System One Bench: Typed decisions (English)	74.5% ✎	▁▁·▁▁▁▅▄▇██▁ beats 25/29	medium · avg 63%	▇ 2.4%	blunt · at its level
LeanMCP · aegis2 response	82.2% ✎	█··▃·▃·█··▅▅ beats 11/11	easy · avg 70%	▇ 2.1%	typical
LeanMCP · helpsteer2 helpfulness	1.0050 ordinal levels ✎	▄▄▄▄▄·▄·█▄▄▄ beats 5/10	avg 1.357 ordinal levels	▇ 1.6%	blunt
LeanMCP · sst5	0.5330 ordinal levels ✎	▂·····▂▂▄█▄▂ beats 6/11	avg 0.879 ordinal levels	▇ 1.6%	blunt
LeanMCP · mcpbench tool selection	100.0% ✎	▂··········█ beats 8/9	very easy · avg 89%	▇ 1.5%	sharp · maxed out · noisy

10 more tests

test	result	vs the others in its run	difficulty	counts	why
LeanMCP · atbench500	75.8% ✎	▃··▅▅█··▅··▅ beats 5/11	medium · avg 63%	▇ 1.4%	sharp · noisy
LeanMCP · prompt injections	74.1% ✎	▂▂▂▂█·▄▂···▂ beats 7/11	medium · avg 62%	▇ 1.1%	typical
LeanMCP · mcpmark tool selection	100.0% ✎	▂··········█ beats 10/11	very easy · avg 96%	▇ 0.7%	sharp · maxed out · noisy
Perch semantic lint · private (Youden J)	0.3944 J ✎	█····▄█▄···█ beats 3/7	avg 0.380 J	▇ 0.4%	blunt
Perch semantic lint · public (Youden J)	0.4019 J ✎	█····▄█▄···█ beats 4/7	avg 0.363 J	▇ 0.4%	blunt
Perch security detection · public (Youden J)	0.1153 J ✎	█···▄▄···▄▄█ beats 7/7	avg 0.052 J	▇ 0.1%	blunt
Perch bug detection · private (Youden J)	0.0241 J ✎	▅··▃▃█·····▃ beats 2/7	avg 0.027 J	▇ 0.1%	blunt
Perch bug detection · public (Youden J)	0.0182 J ✎	█▄·█·▄▄····▄ beats 2/7	avg 0.026 J	▇ 0.1%	blunt
Perch security detection · private (Youden J)	0.0530 J ✎	█▃▃··▃·▃···▃ beats 6/7	avg 0.017 J	▇ 0.1%	blunt
LeanMCP · jailbreak classification	97.0% ✎	▂·▄··▂···█▆▂ beats 10/11	very easy · avg 95%	▇ 0.0%	blunt · maxed out · noisy

$ where to run it

Through an API

api · OpenRouter · price unknown · openrouter.ai ↗

Model repository and documentation: openrouter.ai ↗

$ how the score was worked out · LeanMCP · JevBench 39%, Etsabary · decision benchmarks 20%, System One Bench (Ali Alhashi… 17% and 4 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

source	results	share of the score	licence
LeanMCP · JevBench	18	▇▇▇▇▇▇▇▇▇▇▇▇ 39%	MIT
Etsabary · decision benchmarks	2	▇▇▇▇▇▇ 20%	MIT
System One Bench (Ali Alhashimi)	16	▇▇▇▇▇ 17%	CC-BY-4.0
Decision Models as Judges for Hallucination Detection	5	▇▇▇▇▇ 16%	CC-BY-4.0
Do We Need an LLM to Know Why Sources Conflict? Decision Models for Knowledge-Conflict Typing in RAG	1	▇ 3%	CC-BY-4.0
Freyer et al. · MedQA	1	▇ 3%	CC BY 4.0
Perch benchmark results	6	▇ 1%	CC BY 4.0
prior · pull toward the middle		▇ 1%
weighted average		= 64.5

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit 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 2026-11-03 02:37 UTC · sources and licences · json