An open-weights System One model from Convai Innovations. A fully fine-tuned ModernBERT-large encoder with a from-scratch decision head that scores one marker per option and answers every question in a single 33–39 ms pass. Runs on your own CPU or GPU.
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, route
choice, score, noul, classify, route
Architecture
laya
vega
Fine-tuned from
answerdotai/modernbert-large
qwen/qwen3.5-0.8b
License
apache-2.0
apache-2.0
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
—
55.6%
Calibration error
—
0.151
Valid action rate
—
—
Median latency
39.5 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 laya and vega?
laya is from Convai Innovations and vega from Nandakishor M. Both have open weights you can download and run. Both answer choice, score, noul, classify and route questions. vega reads up to 72K tokens of state, against 512 tokens for laya. laya is the smaller model, at 421M parameters to 800M.
Which is more accurate, laya or vega?
Only vega publishes an accuracy figure (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); laya does not, so there is no comparison to make without your own test.
Which is cheaper, laya or vega?
laya: Free (open weights). vega: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run laya or vega locally?
Yes, both: systemone pull convai-innovations/laya and systemone pull nandakishorm/vega download the weights.
—
Evaluation suite
—
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
