Contrastive Language Models score a state against a set of candidate actions with a contrastive objective. Two projection heads on frozen Qwen3-8B embeddings, trained with InfoNCE; clm-serve maps Choice, Score and Noul onto candidate ranking.
A 2.8 MB byte-level transformer for GUI form filling. For each form element it returns one probability per typed option (fill an entity, check, click or skip), using jevlike's option-attention head.
Decides
choice, score, noul, rank, classify, route
choice
Architecture
clm
cua-s1
Fine-tuned from
qwen/qwen3-8b
—
License
apache-2.0
mit
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
—
100.0%
Calibration error
—
—
Valid action rate
—
—
Median latency
—
—
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 clm and cua-s1-forms?
clm is from Contrastive-LM and cua-s1-forms from Cua. Both have open weights you can download and run. Both answer choice questions. Only clm answers score, noul, rank, classify and route. cua-s1-forms is the smaller model, at 706K parameters to 8.0B. clm is licensed apache-2.0; cua-s1-forms, mit.
Which is more accurate, clm or cua-s1-forms?
Only cua-s1-forms publishes an accuracy figure (100.0% on 196-decision evaluation on real forms); clm does not, so there is no comparison to make without your own test.
Which is cheaper, clm or cua-s1-forms?
clm: Free (open weights). cua-s1-forms: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run clm or cua-s1-forms locally?
Yes, both: systemone pull contrastive-lm/clm and systemone pull cua/cua-s1-forms download the weights.
