A Situated World Model for fast judgment in agent loops.
Juspay's open decision model. Qwen3.6-35B-A3B with a merged rank-16 LoRA, shipped with a pinned SGLang serving bundle that answers Choice, Score and Noul questions over text, images or a video through a TypeSafe-compatible /v1/systemone API.
Decides
choice, route, classify
choice, score, noul, classify, route
Architecture
trio-spark
xor
Fine-tuned from
—
qwen/qwen3.6-35b-a3b
License
proprietary
apache-2.0
Availability
Hosted API
Open weights
Hosted by
MachineFi
—
Input price
$0.042/MTok
—
Decision accuracy
—
90.0%
Calibration error
—
0.073
Valid action rate
—
—
Median latency
—
69 ms
p95 latency
—
162 ms
Evaluation suite
—
JevBench public set (231 items), maker's self-run of Xor 1.2
Latest version
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 trio-spark-v1 and xor?
trio-spark-v1 is from MachineFi and xor from Juspay. trio-spark-v1 is only available as a hosted API; xor has open weights you can download and run. Both answer choice, route and classify questions. Only xor answers score and noul. trio-spark-v1 is licensed proprietary; xor, apache-2.0.
Which is more accurate, trio-spark-v1 or xor?
Only xor publishes an accuracy figure (90.0% on JevBench public set (231 items), maker's self-run of Xor 1.2); trio-spark-v1 does not, so there is no comparison to make without your own test.
Which is cheaper, trio-spark-v1 or xor?
trio-spark-v1: $0.042 / $0 per 1M. xor: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run trio-spark-v1 or xor locally?
xor yes — systemone pull juspay/xor downloads its weights. The other is only served as a hosted API.
