Together AI's Jev-style classifier. A LoRA fine-tune of Qwen3.5-4B that reads a state, a question and 2 to 24 options and returns one option letter. Served on Together's platform; recipe published.
A Situated World Model for fast judgment in agent loops.
Decides
choice, classify, route
choice, route, classify
Architecture
tev
trio-spark
Fine-tuned from
qwen/qwen3.5-4b
—
License
Unspecified — weights licence being finalised
proprietary
Availability
Open weights + hosted API
Hosted API
Hosted by
Together AI
MachineFi
Input price
$0.042/MTok
$0.042/MTok
Decision accuracy
88.0%
—
Calibration error
—
—
Valid action rate
—
—
Median latency
—
—
p95 latency
—
—
Evaluation suite
Together development set (reused, not held out)
—
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 tev1 and trio-spark-v1?
tev1 is from Together AI and trio-spark-v1 from MachineFi. tev1 has open weights and a hosted API; trio-spark-v1 is only available as a hosted API. Both answer choice, classify and route questions. tev1 is licensed other; trio-spark-v1, proprietary.
Which is more accurate, tev1 or trio-spark-v1?
Only tev1 publishes an accuracy figure (88.0% on Together development set (reused, not held out)); trio-spark-v1 does not, so there is no comparison to make without your own test.
Which is cheaper, tev1 or trio-spark-v1?
tev1: $0.042 / $0 per 1M, or free to self-host. trio-spark-v1: $0.042 / $0 per 1M. Open weights cost nothing per call beyond your own hardware.
Can I run tev1 or trio-spark-v1 locally?
tev1 yes — systemone pull together-ai/tev1 downloads its weights. The other is only served as a hosted API.
