vllm-sr/vela3-turbo

Name it. Vela finds it.
_{vLLM Semantic Router × KR Labs · Vela 3.0 family · vela3-turbo · 27B · base d3-turbo · Apache-2.0}

What it does · Quick start · Evaluation · The family · 📡 Collection

---

vela3-turbo is the 27B member of the Vela 3.0 family, built on d3-turbo. In one system_one call it answers routing and general decisions and finds the exact spans you name (entities, relations, personal data, unsupported claims, evidence), with probabilities and character offsets and no generated text.

Parameters	26.09B with vision encoder (base) + 5044 MB of adapters and heads
Base	vllm-sr/d3-turbo v1.0, revision da17b5d0c860f7e5be6e2afebe67228b4bd24d0c, unchanged (the public d3-turbo main is now v1.1; this package stays on v1.0)
Inputs	Text or JSON, plus images and videos for d3-turbo's own question types
Answers	Choice · Yes / No · Score · Span
Licence	Apache-2.0

Summary

vela3-turbo is the reference model of the Vela 3.0 family, on d3-turbo v1.0 (pinned). d3-turbo, frozen, answers every ordinary decision, bit for bit, and gated adapters add the spans you name, entities, relations and router signals. It has the family's top scores on classification and relations.

Highlights

• Zero-shot classification: 81.4 on the 13-set GLiFormer classification suite, above every model the paper reports (GPT-5-mini: 79.8, Gemma-4-31B-IT: 79.69; reported, GLiFormer paper Tab 5).
• Relations: 35.9 end to end (GLiFormer Table 4 protocol, avg of 4), above every model reported (best: Qwen3.5-9B, prompted, 25.7; reported, GLiFormer paper Tab 4); 51.7 on CoNLL04 with the head entity given (GLiFormer-base: 8.6, our run, same harness).
• Zero-shot NER: 64.6 on OOD-7 (CrossNER + MIT) from label names, above every GLiNER / GLiFormer number reported on it (best: gliner-multitask-large-v0.5, 62.76).
• Unsupported claims: 0.900 example-F1 on the 10,698-answer LettuceDetect v2 test set, above LettuceDetect v2 mmBERT-base: 0.869 (reported, its card) and Vela 2.0 9B: 0.885 (its card).
• Router signals: better than Vela 2.0 9B on 20 of the 24 test files of the public router signal suite, level on 4, worse on none (paired 95% CI).
• Decisions, bit for bit: every ordinary decision is answered by d3-turbo@da17b5d0 itself, adapters off (561 of 561 identical in our check).

_{vela3-turbo (27B) against reported numbers. Accent = vela3-turbo; grey = other models (reported).}

At a glance

What it is	A 26B decision-and-extraction model for LLM routing and guardrails. It runs vllm-sr/d3-turbo, the largest d3 decision model, pinned to d3-turbo v1.0 (revision da17b5d0; the public main is now v1.1), unchanged, and adds the exact spans you name: entities, relations, personal data, unsupported claims and evidence.
How it was trained	d3-turbo stays frozen, so every ordinary decision is d3-turbo's own answer, bit for bit. Vela adds small gated LoRA adapters on top, one per job: spans (unsupported claims, personal data, evidence), entities, relations and router signals. Each adapter switches on only for its own questions. The adapters were trained on vela3-flash's recipe and ported to d3-turbo: span, entity and router adapters re-thresholded, the relation adapter retrained. Router signals are trained supervised, as in Vela 2.0, on the public router datasets Vela 2.0 used (for example AEGIS 2.0, PolyGuardMix and Nemotron-Safety-Guard v3 for safety, Global-MMLU for domain, WildFeedback for feedback, DiffusionDB for modality, MASSIVE for intents); domain and feedback also on the public router signal suite's train split. Other training data: public NER, relation, hallucination, evidence and PII datasets, plus web and document text labelled by a large teacher model. No test split and no Decision Index item is trained on; some router-signal-suite test rows also occur in public training sources, so the router table also reports the rows without that overlap.
What it is good at	Zero-shot classification; zero-shot relations; decisions, bit for bit; router signals above Vela 2.0 9B on 7 of 8 signals of the public router signal suite. Pick it for the most accurate answers, as the family's reference.
How we measure it	Public test sets only, run through this package one request at a time with the shipped thresholds: CoNLL04, CoNLL04 · DocRED · FewRel · CrossRE (avg), OOD-7 (CrossNER + MIT), RAGTruth, Super-NI task 1568, the 13-set GLiFormer classification suite, the LettuceDetect v2 test set, the public router signal suite, the Decision Index 0.4.1. Other models' numbers are as their authors report them on the same test set and metric, or scored by us on the same rows where a table says so.

What it does

Capability	Ask (one question of a system_one call)	Get
Decisions	{"type": "choice", "instructions": "Which topic?", "criteria": {"billing": None, "tech": None}} (also noul, score)	a probability per option, from d3-turbo itself
Router signals	a built-in router question id from router_questions.json	the signal's answer from the gated decision adapter
Named entities	{"type": "span", "criteria": ["drug", "dosage"]}	spans with label, offsets and score
Relations	{"type": "span", "instructions": "Which spans are the founder of Apple?", "criteria": {"founder": None}}	the related spans
Personal data	{"type": "span", "criteria": ["PERSON", "EMAIL_ADDRESS"]}	spans of the 17 trained PII types, any subset
Unsupported claims	state {"context": ..., "answer": ...} and {"type": "span", "criteria": {"unsupported": None}}	the words of the answer the context does not support
Evidence	{"type": "span", "criteria": ["evidence"]}	the sentences that support the answer
Boxes, tracks, images in questions	d3-turbo's box / track questions, assets	answered by d3-turbo itself, adapters off (d3-turbo card)

Quick start

# pip install "transformers==5.17.0" torch torchvision pillow opencv-python-headless safetensors accelerate
from transformers import AutoModel

m = AutoModel.from_pretrained("vllm-sr/vela3-turbo", trust_remote_code=True)  # downloads d3-turbo at the pinned revision

# Name any labels; the spans come back with offsets and scores.
r = m.system_one(
    state="Geoffrey Hinton presented capsule networks at NeurIPS in Montreal.",
    questions={"ents": {"type": "span", "criteria": ["researcher", "algorithm", "conference", "location"]}})
# researcher "Geoffrey Hinton", algorithm "capsule networks", conference "NeurIPS", location "Montreal"

# Unsupported claims in an answer, against a context or an image.
r = m.system_one(
    state={"context": "The Eiffel Tower is 330 metres tall and was completed in 1889.",
           "answer": "The Eiffel Tower is 330 metres tall and was completed in 1925 by Gustave Eiffel's son."},
    questions={"halu": {"type": "span", "criteria": {"unsupported": None}}})
# flags "1925 by Gustave Eiffel's son"; add images=["photo.jpg"] to check an answer against a photo

# PII, relations and evidence use the same call.
r = m.system_one(
    state="Steve Jobs founded Apple in Cupertino together with Steve Wozniak.",
    questions={"pii": {"type": "span", "criteria": ["PERSON"]},
               "rel": {"type": "span", "instructions": "Which spans are the founder of Apple?", "criteria": {"founder": None}}})

Evaluation

All numbers on public test sets, measured through this package with system_one, one request at a time and the shipped thresholds. Other models' numbers are as reported by their authors (or scored in the same harness where marked). Full per-set tables: EVALUATION.md.

Named entities

Benchmark	vela3-turbo	Best other model (reported)	Vela 2.0 9B	Size-class best (>10B, reported)
OOD-7, label names only (default; F1)	64.6	B2NER-20B 72.1 (standardised label names; B2NER Tab 3)	43.5	B2NER-20B, standardised label names 72.1 (B2NER, Table 3)
NER, 26 datasets of the GLiFormer paper (label names; not all zero-shot)	63.7	–	–	–

Label names are the only input. Training holds no CrossNER or MIT text, though many generic CrossNER type names occur as training labels.

Relations

Benchmark	vela3-turbo	Best other model (reported)	Vela 2.0 9B	Size-class best (>10B, reported)
CoNLL04, head entity given, label names (F1)	51.7	GLiFormer-base 8.6 (same harness, Vela 2.0 card)	10.1	–
End to end, GLiFormer Table 4 protocol, avg of 4 (strict F1)	35.9	Qwen3.5-9B, prompted 25.7 (GLiFormer paper Tab 4)	–	Gemma-4-31B-IT, prompted 25.1
End to end: CoNLL04 · DocRED · FewRel · CrossRE (CoNLL04 zero-shot, others in-domain)	42.8 · 33.7 · 27.8 · 39.2	GPT-5-mini 42.4 · GLiNER-Relex 31.3 · Qwen3.5-9B 71.6 · GLiNER-Relex 18.1 (GLiFormer paper Tab 4)	–	–

CrossRE 39.2 (in-domain) is above every model reported in the GLiFormer paper (best: GLiNER-Relex 18.1).

Evidence

Benchmark	vela3-turbo	Best other model (reported)	Vela 2.0 9B	Size-class best (>10B, reported)
QASPER, word-F1 (zero-shot)	39.3	–	–	–
ACL-Verbatim, word-F1 (in-domain)	49.1	ACL-Verbatim ModernBERT 53.4 (in-domain, evidence only; Vela 2.0 card)	24.5	–

ACL-Verbatim is in-domain: its silver training split is in the training data.

Unsupported claims (hallucination spans)

Benchmark	vela3-turbo	Best other model (reported)	Vela 2.0 9B	Size-class best (>10B, reported)
LettuceDetect v2 test set, 10,698 answers (example-F1)	0.900	LettuceDetect v2 qwen-2b 0.921 (its card)	0.885	–
RAGTruth (example-F1)	0.779	HDM-2 (3B) 0.850 (HalluciNot Tab 2)	0.774	Llama-2-13B fine-tuned 0.787 (reported, RAGTruth Tab 5)

Zero-shot classification

Benchmark	vela3-turbo	Best other model (reported)	Vela 2.0 9B	Size-class best (>10B, reported)
GLiFormer suite, 13 sets, label names only (mean macro-F1)	81.4	GPT-5-mini 79.8 (GLiFormer paper Tab 5)	–	Gemma-4-31B-IT 79.7

These questions run on d3-turbo with every adapter off.

Router signals (public router signal suite; above Vela 2.0 9B on 7 of 8)

Signal (metric, test rows)	vela3-turbo	Vela 2.0 9B, same rows	Decision-1.0-Route-0.6B, same rows	Rows sharing no text with training: vela3-turbo / Vela 2.0 9B
Harmful request (AUC, 2,230)	0.958	0.957	0.963	0.953 / 0.958 (1,254)
Jailbreak or prompt injection (AUC, 1,878)	0.967	0.919	0.991	0.965 / 0.902 (944)
Personal data in the request (AUC, 3,000)	0.971	0.960	0.994	0.973 / 0.960 (2,802)
Answer not supported by its context (AUC, 3,000)	0.783	0.765	0.791	0.727 / 0.744 (1,743)
Fact-check needed (AUC, 800)	0.870	0.809	0.942	0.863 / 0.805 (690)
Output modality (text or image) (AUC, 3,000)	0.993	0.933	0.999	0.993 / 0.930 (2,916)
Domain (subject area) (accuracy, 2,996)	0.884	0.765	0.863	0.885 / 0.763 (2,916)
User feedback on the previous answer (accuracy, 2,842)	0.852	0.666	0.848	0.843 / 0.631 (2,295)

vllm-sr/router-signal-suite (public), the test file of each signal, scored with the dataset's own scoring code. Each signal is a router question this package was trained on (supervised, as in Vela 2.0), asked one request at a time with the shipped thresholds.

Vela 2.0 9B was run by us through its public release on the same rows. Decision-1.0-Route-0.6B is scored from the suite's published predictions; it was trained on the suite's own train split, so these test files are in-distribution for it.

Bold: the paired 95% bootstrap CI against Vela 2.0 excludes a tie. The last column drops the test rows that share text (exact or a 13-gram) with this model's training data. Safety and jailbreak are scored on the rows whose text the suite publishes (2,230 of 3,003 and 1,878 of 3,000).

Domain and feedback are also trained on the suite's own train split (with the public router data Vela 2.0 used; the suite's test and hold rows, and anything sharing an exact text or a 13-gram with them, are kept out), as Decision-1.0-Route is, so their test files are in-distribution for both. The domain question has a 15th option, religion, which the suite maps to philosophy. Feedback has NO_FEEDBACK as a trained option.

Decisions (general decisions are the base model's, bit for bit)

Decision Index 0.4.1 (2026-10-21)	Public index	Full score (board rank)
Fastino GLiDE V1.2	71.99	71.52 (1)
Perplexity Decider v1.2 (27B) (tied rank)	67.74	67.98 (4)
vLLM-SR d3-turbo v1.1 (27B) (the public d3-turbo main; not this package's base; tied rank)	71.62	67.31 (10)
vLLM-SR d3 (27B) (tied rank)	65.45	67.15 (11)
GLiClass v4 Edge (0.4B) (board data of 2026-10-22T20:10Z, which added it after this snapshot; tied rank)	47.49	31.21 (152)
OneJev1.6-0.8B (tied rank)	48.76	29.58 (155)
StartLux-Decision v1.1 0.8B (tied rank)	53.79	28.88 (158)
General decisions = d3-turbo (d3-turbo v1.0, revision da17b5d0: not on the 0.4.1 board)	–	–

General decisions are the base model's own answers, bit for bit; no Vela package was run on the board. Full score (the board's ranking): 60% private same-skill tests and 40% private new-domain tasks; public index: the 37 public benchmarks, chance-corrected. All values reported on the Decision Index 0.4.1 board (2026-10-21; multimodalart/jev-decision-index, data generated 2026-10-22) unless noted; tied ranks share a number. d3-turbo v1.0, this package's base (revision da17b5d0), is not on the Decision Index 0.4.1 board; the public d3-turbo main is v1.1, which is on the board at 67.31 (rank 10, tied).

Benchmark	vela3-turbo	Best other model (reported)	Vela 2.0 9B	Size-class best (>10B, reported)
Super-NI task 1568, held out (accuracy)	0.636	–	0.574	–
MASSIVE, three unseen scenarios: intent / scenario (accuracy; 14,911 decontaminated rows)	0.833 / 0.937	–	0.763 / 0.871 (full set)	–

MASSIVE: three scenarios never used in our router training (iot, transport, cooking), 52 locales; accuracy on the 14,911 test utterances that match no training text of the model's lineage. The 342 excluded utterances occur verbatim in upstream training mixtures (tasksource MASSIVE / HWU64 rows). Vela 2.0 values are on the full 15,253, so these rows are not bolded. Super-NI 1568: 79 to 90 of its 162 ProPara passages share a 13-gram with other tasks' training rows.

Latency

Median, one request, AMD Instinct MI325X (measured alone on one GPU, no other process on it)	200 tokens	1K tokens	3K tokens
4 router decisions	469 ms	952 ms	2238 ms
1 span question	84 ms	182 ms	531 ms
4 decisions + 1 span	566 ms	1138 ms	2768 ms

Span questions with label names only.

Benchmarks

• OOD-7 (CrossNER AI, literature, music, politics, science; MIT movie, restaurant): find the entities of the named types in text from domains the model was not trained on (strict entity F1).
• NER, 26 datasets of the GLiFormer paper (IE_INSTRUCTIONS): entities of named types across 26 public NER sets, not all zero-shot (strict entity F1).
• CoNLL04, head entity given: given one entity of a sentence, find the entities it is related to by the named relation (F1 over those tail entities).
• Relations end to end, GLiFormer Table 4 protocol (CoNLL04, DocRED, FewRel, CrossRE): find the entity pairs and their relation from raw text (strict triplet F1).
• QASPER: find the sentences of a research paper that answer a question about it (word-F1).
• ACL-Verbatim (KRLabsOrg/acl-verbatim-spans): find the passage of an ACL paper that supports an answer (word-F1).
• LettuceDetect v2 test set (10,698 answers, as in the LettuceDetect v2 paper (GroundLM 2026 workshop at EMNLP 2026) and card): does an answer contain claims its context does not support (example-F1: the answer is flagged or not).
• RAGTruth: does a retrieval-augmented answer contain claims its retrieved context does not support (example-F1).
• MASSIVE, three unseen scenarios (iot, transport, cooking; 52 languages): pick the intent and the scenario of a user command the model never saw in training (accuracy).
• Super-NI task 1568 (ProPara): answer a multiple-choice question about how a process changes an entity (accuracy).
• Zero-shot classification, 13 sets (the GLiFormer paper's suite: SST-2, IMDB, AG News, Banking77, and more): pick the label of a text from label names only (mean macro-F1).
• Decision Index 0.4.1 (board, 2026-10-21): 37 public decision benchmarks plus private tests; the Full score ranks on the private same-skill and new-domain tests (chance-corrected skill, 0 = guessing).
• Router signal suite (vllm-sr/router-signal-suite): the vLLM Semantic Router's public test of its request signals (harmful requests, jailbreaks, personal data, unsupported answers, fact-check need, modality, domain, feedback); each label comes from public corpora (AUC or accuracy).

The family

Model	Size	Base	Best for
vela3-fast	0.4B	UltraBERT-380M encoder	the default router; fastest: decisions + spans in one ~14 ms encoder pass, English, 32K
vela3-fast-vision	0.25B	UltraVBERT encoder	image decisions, image-grounded hallucination spans, region pointing (preview)
vela3-lite	0.8B	d3-lite	small GPUs, low latency
vela3-mini	4B	d3-mini	balanced accuracy and speed
vela3-flash	9B	d3-flash	high-accuracy routing
vela3-turbo	27B	d3-turbo	the reference model

How it works

• Frozen base. d3-turbo@da17b5d0 runs unchanged; ordinary questions are its own answers.
• Gated LoRA adapters. Each switches on only for its own questions, and every answer names the one it used (head_reason): decision (router signals), NER (entity labels with definitions), relation, image (hallucination spans with images), span (evidence, PII, unsupported claims, toxic spans, names-only labels).
• Spans as decisions. The labels are presented as options; a span head scores every word against every label, and agreeing neighbours form a span with character offsets.
• Label names in. Labels are read by name at this size; no definitions are written.
• Specification. Adds 5044 MB of adapters and heads; text or JSON state; requests with images, videos, assets, box or track questions, or long_context go to the base with every adapter off. Full evaluation: EVALUATION.md.

Trade-offs

• Unsupported claims: hallucination spans 0.900 against 0.921 for LettuceDetect v2 qwen-2b, RAGTruth 0.779 against 0.850 for HDM-2 (3B) (reported).
• NER from label names: 64.6 on OOD-7, above the best reported encoder (gliner-multitask-v1.0 63.25) and below B2NER-20B, standardised label names 72.1 (reported).
• End-to-end relations on generic types: 35.9 average on the four typed sets, above every model in GLiFormer Table 4 (reported), but 0.2 on SciERC and 0.5 on NYT11, where every entity has the generic type entity.

Citation

@misc{vela3_turbo,
  title        = {vela3-turbo (Vela 3.0, 27B): zero-shot extraction and router signals on Decision 3.0},
  author       = {{KR Labs} and {vLLM Semantic Router}},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/vllm-sr/vela3-turbo}}
}