JPT-9B

JPT-0.8B · JPT-4B · JPT-9B · JPT-35B-A3B · llm2jev

What is JPT-9B

JPT-9B is a fast, open decision model: give it a situation and typed questions, get a calibrated probability for every option from one forward pass. No generated explanation, no reasoning tokens — latency is one prefill.

The large sibling of JPT-4B: the same recipe and the same data.

It implements the typed-decision interface introduced by Jev from TypeSafe AI [1]: a caller sends a state plus questions, and each question is one of three types. JPT is an independent model, not derived from Jev and not trained on Jev outputs; it is an open alternative behind the same interface.

Question type	What it answers	Options
choice	pick one	2–255 labels
score	a level on an ordered scale	the scale's levels
noul	yes / no	true, false

Built on Qwen/Qwen3.5-9B: a LoRA fine-tune merged into full weights. The vision tower is unchanged.

Probabilities use one temperature T = 1.087, fit once on a held-out split — never per benchmark.

❯❯ Benchmarks

❯ JevBench v1.4.0

JevBench [2] scores general typed decisions; JPT-9B reaches 0.853 public accuracy, level with JevK5 and just under Winnow-12B and Jev 1.13.0 — but below the smaller JPT-4B (0.879). v1.4.0 has 231 public items and 308 sealed items only the maintainer can run, so JPT-9B has no official v1.4 score yet.

Table

System	Params	Public accuracy (231)	Sealed accuracy (308)	v1.4 score
JPT-35B-A3B (ours)	35B-A3B	0.892	pending	pending
JPT-4B (ours)	4B	0.879	pending	pending
Jev 1.13.0 (TypeSafe AI, API)	closed	0.866	0.367	63.3
Winnow-12B Q8	12B	0.857	0.331	55.6
JPT-9B	9B	0.853	pending	pending
JevK5 v0.2.0	27B	0.853	0.331	62.0
decider-35b-a3b	35B-A3B	0.831	0.315	41.2
Hopper	—	0.823	0.341	59.4
openjev 4B v5¹	4B	0.814	—	—
SemIf, formerly OpenJev (Qwen3.5-4B)	4B	0.810	0.263	47.7
local-jev Qwen3.5-4B	4B	0.805	0.260	46.8
reflex 4B	4B	0.792	0.282	54.0
kev 4B (research preview)	4B	0.662	0.224	36.1

Source: other rows from results/v1.4/jevbench-v1.4-results.json at jevbench commit 2fa63fa (2026-09-23). Ours measured with JevBench's CLI at 2fa63fa (SGLang 0.5.18, kirp/jpt-9b@7114b0c): 197/231, easy 48 · standard 68 · hard 81. Our own harness gets 198: on original-intent-02-0 two options tie exactly (0.4864 each) and the two tools break the tie differently. ¹ Not in the v1.4 results; its own card's number on the 231 public items. A same-prompt Qwen3.5-9B zero-shot run on the public items has not been done yet.

❯ Decision Index 0.2.1

Decision Index [3] 0.2.1 (2026-09-27) is the broadest test: the full frozen suite, 38 scored benchmarks in five areas, chance-corrected — JPT-9B scores 46.89, the best 9B on the board and 0.2 behind Decider 35B-A3B. Run through llm2jev over SGLang and submitted as apolinario/decision-index#7; 0.2.1 rescores the same run (42.73 under 0.2).

Table

Model	Params	Decision Index 0.2.1
Jev 1.13.0 (TypeSafe AI, API)	closed	57.91
JPT-35B-A3B (ours, not on the board yet)	35B-A3B	52.89
Decider 35B-A3B	35B-A3B	47.11
JPT-9B	9B	46.89
Decision 1.0 Lux	9B	43.49
Bespoke Nimble 9B v2	9B	39.57
Kev 9B	9B	38.48

Source: live board data/index-v0.2.1.json (generated 2026-09-27 16:59 UTC); full run and scores.json in kirp/decision-index-results-jpt-9b (gated: it carries the suite's GPQA/HLE item text).

By area, against Jev 1.13.0 on the same items and scorer: JPT-9B is behind Jev in all five areas and ahead of it on 4 of the 38 index benchmarks.

Per-area and per-benchmark skill vs Jev 1.13.0

Area	Jev 1.13.0	JPT-9B
Knowledge	51.4	31.7
Language	62.0	56.7
Retrieval	55.4	44.6
Tools	75.1	67.0
Arts	37.7	28.6

Area	Benchmark	Jev 1.13.0	JPT-9B
Arts	BPoMP	81.8	72.4
Arts	ForecastBench	30.6	27.4
Arts	Habermas Machine	21.5	13.2
Arts	Humicroedit	23.7	19.5
Arts	New Yorker	62.6	52.6
Arts	POP909-CL	15.9	3.5
Arts	cfcolor	28.8	11.4
Games	ChessBench	9.8	7.3
Knowledge	BBH	89.7	55.6
Knowledge	CLadder	45.3	33.8
Knowledge	CRUXEval	57.1	29.6
Knowledge	GPQA Diamond	71.4	21.8
Knowledge	GSM8K	75.6	69.2
Knowledge	HLE	4.7	0.0
Knowledge	MMLU-Pro	80.5	49.1
Knowledge	MuSR	46.1	28.1
Knowledge	SATA-Bench	25.4	22.6
Language	ACOS	27.3	20.3
Language	ANLI	62.2	50.0
Language	ContractNLI	59.1	74.2
Language	FinEntity	80.8	90.0
Language	HellaSwag	92.7	81.2
Language	NLI4CT	69.0	59.1
Language	RAGTruth	51.3	50.6
Language	VAST	46.9	46.6
Language	WinoGrande	83.9	49.2
Language	iSarcasmEval	36.3	43.9
Retrieval	Amazon ESCI	43.8	40.6
Retrieval	BANKING77	79.5	65.0
Retrieval	BRIGHT	40.6	39.9
Retrieval	CLINC150+OOS	89.2	35.8
Retrieval	HoVer	45.7	32.7
Retrieval	PhishNChips phishing decisions	25.1	52.1
Tools	API-Bank	88.0	78.3
Tools	BFCL	94.3	93.1
Tools	Home appliance simulator	52.3	43.2
Tools	ToolRet	59.9	55.8
Tools	When2Call	74.6	57.3

Chance-corrected skill × 100 (0 = random, 100 = perfect). Jev's numbers are its official entry on the live board (jev-1.13.0); ours are from the same kit and suite.

❯ More benchmarks

Against JPT-4B on the same evals: JPT-9B is better on NLI, intent classification and EnvBench public, and worse on the JevBench hard tier. Same-prompt base-model (Qwen3.5-9B zero-shot) rows exist only for EnvBench; the image evals (ScreenSpot-v2, Screen2Words, ERQA) have not been run on this checkpoint.

Benchmark (version, n)	What it tests	JPT-9B	JPT-4B	Jev 1.13.0
JevBench v1.4.0 public hard tier [2] (111)	hardest general decisions	0.730 (ECE 0.097, Brier 0.382)	0.784	—
Typed decisions test (ours, 2,000)	in-distribution typed decisions	0.806 (ECE 0.149, Brier 0.316)	0.796	—
ANLI r1 / r2 / r3 [4] (dev)	adversarial NLI	0.737 / 0.647 / 0.677	0.697 / — / 0.613	—
Banking77 [5] / MASSIVE 1.1 [6] (en / de / zh)	intent classification	0.787 / 0.897 / 0.853 / 0.830	0.757 / 0.857 / 0.833 / 0.837	—
AG News [7] / Emotion [8] / SST-5 [9] (test)	text classification (train splits in the mix)	0.917 / 0.557 / 0.603	—	—
EnvBench v0.1 (ours) public / held-out [10] (skill 0–100)	sequential decisions in game envs	50.5 / 46.8	47.7 / 47.0	—

Jev 1.13.0 has no official score on these splits (our own test/dev cuts, EnvBench, and the image sets), so its column is "—"; its official scores on the Decision Index versions of ANLI and BANKING77 are in the per-benchmark table above. Running the Jev API on these splits would fill them.

Qwen3.5-9B, same prompt, zero-shot on EnvBench v0.1: 24.2 public / 25.3 held-out. Banking77, MASSIVE, AG News, Emotion, SST-5 and typed rows are in-distribution (train splits in the mix, test items not).

❯❯ Quick Start

Two pieces: an engine that holds the weights, and llm2jev (>= 0.6.1) in front of it, reading option probabilities off the engine.

⚡ SGLang (recommended)

python -m sglang.launch_server --model-path kirp/jpt-9b --port 30000 \
  --context-length 32768 --mamba-scheduler-strategy extra_buffer &  # Qwen3.5's DeltaNet layers need this flag
llm2jev --model kirp/jpt-9b --backend sglang --url http://127.0.0.1:30000 --port 8080 --temperature 1.087

Tested with SGLang 0.5.9; install cuDNN 9.15+ over its pinned 9.10: pip install "sglang==0.5.9" && pip install "nvidia-cudnn-cu12>=9.15".

🔁 vLLM

vllm serve kirp/jpt-9b --max-logprobs 256 --return-tokens-as-token-ids --enable-scale-out --port 8000
llm2jev --model kirp/jpt-9b --backend vllm --url http://127.0.0.1:8000 --port 8080 --temperature 1.087

The three vLLM flags are required: without them every request is a bare HTTP 400.

🧪 No engine (quick check only)

pip install "llm2jev[hf,vision]"
llm2jev --model kirp/jpt-9b --backend hf --port 8080 --temperature 1.087

Serializes requests; for traffic use SGLang or vLLM.

📨 Ask it a question

import requests
r = requests.post("http://127.0.0.1:8080/v1/systemone", json={
    "state": "Refund policy: full refund within 30 days of purchase; 50% until day 60; none after.\n"
             "Order 1182 was bought on 3 March and returned on 20 April.",
    "questions": {
        "refund": {"type": "choice", "instructions": "What refund does order 1182 get?",
                   "criteria": {"full": "Full refund", "half": "50% refund", "none": "No refund"}},
        "late":   {"type": "noul", "instructions": "Was the return made after day 30?",
                   "criteria": {"true": "Yes", "false": "No"}}}})
print(r.json()["answers"])   # each answer has the per-option probabilities

🖼️ With images

state = [{"role": "user", "content": [
    {"type": "image", "image": "https://example.com/screen.png"},
    {"type": "text", "text": "Task: open the settings page. Numbered boxes mark clickable elements."}]}]
questions = {"click": {"type": "choice", "instructions": "Which box should be clicked?",
                       "criteria": {"1": None, "2": None, "3": None, "4": None, "5": None}}}

❯❯ Training

The JPT-4B recipe and data (mix_train_env_v11, 49,221 typed questions) on a larger base, one epoch, merged into full weights.

Part	What it is
Method	LoRA r=16 on every attention, DeltaNet and MLP projection of the language model, lr 5e-5; vision tower untouched
Loss	multi-class Brier over the option labels, on llm2jev's chat prompt with thinking disabled
Batch	8 GPUs × 1 × 5 gradient-accumulation steps = 40 questions per step
Data	49,221 questions in 32,835 records; one epoch over two option-shuffled copies — sources on the JPT-4B card
Held out	no item from JevBench, EnvBench held-out seeds, the Decision Index frozen suite or our typed test split

❯❯ Limitations

• Not strictly better than JPT-4B. The JevBench hard tier (0.730 vs 0.784) and the EnvBench held-out game area (0.288 vs 0.315) are lower, despite lower training and validation loss throughout. Not yet root-caused.
• Arithmetic and dates are its weakest area: it answers from the evidence given and has no reasoning phase by design.
• Up to 255 options are accepted; training covered up to 77 (Banking77).
• English first. Other languages come only from a few multilingual classification sets.
• Images are untested on this checkpoint; they go zero-shot through the base vision tower.

❯❯ References

1. TypeSafe AI. Jev. https://typesafe.ai
2. F. Standhartinger. JevBench, v1.4.0. https://github.com/fstandhartinger/jevbench
3. Decision Index, edition 0.2.1. https://huggingface.co/spaces/multimodalart/jev-decision-index
4. Nie et al. Adversarial NLI. ACL 2020.
5. Casanueva et al. Efficient Intent Detection with Dual Sentence Encoders (Banking77). NLP4ConvAI 2020.
6. FitzGerald et al. MASSIVE. ACL 2023.
7. Zhang et al. Character-level Convolutional Networks for Text Classification (AG News). NeurIPS 2015.
8. Saravia et al. CARER: Contextualized Affect Representations for Emotion Recognition. EMNLP 2018.
9. Socher et al. Recursive Deep Models for Semantic Compositionality (SST). EMNLP 2013.
10. EnvBench, v0.1 (ours, frozen 2026-09-23; not yet public): programmatically solved game, planning and rule decisions with exact gold answers.

❯❯ License

CC BY-NC 4.0. The weights derive from Qwen3.5-9B (Apache-2.0), but some training datasets allow only non-commercial or research use, so the model is released for non-commercial use.

---

JPT-9B is an independent open model that implements a typed-decision interface (noul, choice and score questions answered with probabilities). It is not affiliated with, endorsed by or derived from TypeSafe AI or its Jev model, and it was not trained on Jev outputs.