Fastino's open-weight decision model. A DeBERTa-v3-large encoder that scores label sets supplied at runtime, answers single-label, multi-label, yes/no and ordinal questions in one pass, and extracts spans and relations.
Nandakishor M's typed-decision model. A frozen Qwen3.5 backbone is read, never trained, and cut off mid-stack; a small trained physics engine settles each choice, score or yes/no question into calibrated probabilities, a conformal set and an abstain flag. 0.8B and 4B.
Decides
choice, score, noul, classify, extract, route
choice, score, noul, classify, route
Architecture
gliner2
vega
Fine-tuned from
fastino/gliner2-large-v1
qwen/qwen3.5-0.8b
License
apache-2.0
apache-2.0
Availability
Open weights + hosted API
Open weights
Hosted by
Fastino
—
Input price
—
—
Decision accuracy
60.2%
55.6%
Calibration error
—
0.151
Valid action rate
—
—
Median latency
38.3 ms
—
p95 latency
—
—
Figures are from each model’s manifest; accuracy and latency are what the publishers report, on their own suites and hardware. Add a third model.
Questions
What is the difference between gliner2-5-decide and vega?
gliner2-5-decide is from Fastino Labs and vega from Nandakishor M. gliner2-5-decide has open weights and a hosted API; vega has open weights you can download and run. Both answer choice, score, noul, classify and route questions. Only gliner2-5-decide answers extract. gliner2-5-decide is the smaller model, at 340M parameters to 800M.
Which is more accurate, gliner2-5-decide or vega?
They report on different suites — gliner2-5-decide 60.2% on Fastino fast-decisions suite (17 datasets), vega 55.6% on His 11 October 2026 benchmark, published with its results and its harness (a notebook in vegaml): macro accuracy over 43 public tasks for the 0.8B, 0.717 for the 4B; with JevBench (231 items) and DecisionBench (1,075 rows), 3,841 decisions in all, seed 20261011. On his own set of 2,050 choice decisions over six intent workflows (esci, clinc150, appliances, banking77, when2call, apibank), one choice question each, which he has not published with its question wording, he reports 0.763 (calibration error 0.026) for the 0.8B with its public_intents adapter, 0.389 without the adapter and 0.803 for the 4B with it. He publishes no accuracy for the typed_decisions adapters — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, gliner2-5-decide or vega?
gliner2-5-decide: Hosted, price not published, or free to self-host. vega: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run gliner2-5-decide or vega locally?
Yes, both: systemone pull fastino-labs/gliner2-5-decide and systemone pull nandakishorm/vega download the weights.
Evaluation suite
Fastino fast-decisions suite (17 datasets)
His 11 October 2026 benchmark, published with its results and its harness (a notebook in vegaml): macro accuracy over 43 public tasks for the 0.8B, 0.717 for the 4B; with JevBench (231 items) and DecisionBench (1,075 rows), 3,841 decisions in all, seed 20261011. On his own set of 2,050 choice decisions over six intent workflows (esci, clinc150, appliances, banking77, when2call, apibank), one choice question each, which he has not published with its question wording, he reports 0.763 (calibration error 0.026) for the 0.8B with its public_intents adapter, 0.389 without the adapter and 0.803 for the 4B with it. He publishes no accuracy for the typed_decisions adapters
