Decision modelOpen weights
CLM-v0.1-8B
Contrastive-LM
- Released
- -
- Data date
- September 26, 2026
- Model class
CLM-v0.1-8B takes a different route from autoregressive decision models. A frozen 8B encoder is paired with two trained projection heads and a contrastive learning objective. CLM scores a state against candidates you provide and returns Choice, Noul, Score, or a ranking with relative probabilities. State and action representations can be computed separately, which can speed up repeated candidate scoring.
The model is available as a local download and HTTP interface and returns structured candidate scores. The DeepSWE and Terminal-Bench results belong to fine-tuned verifier heads, not the untouched CLM-v0.1-8B checkpoint.
Specifications and access
| Specification | Value and source |
|---|---|
| Model class | Contrastive System One decision modelSource |
| Model ID | Contrastive-LM/CLM-v0.1-8BSource |
| Input | State, questions, and answer candidatesSource |
| Result | Choice, Noul, Score, and ranking with relative probabilitiesSource |
| Access | Local download and local HTTP interfaceSource |
| License | Apache-2.0Source |
| Benchmark note | The reported DeepSWE and Terminal-Bench scores belong to fine-tuned verifier heads, not the untuned base checkpoint.Source |
Popularity
35.6 / 100
Sources and data date
| Type | Evidence and data date |
|---|---|
| Additional source | CLM model card (retrieved September 26, 2026) · CLM model card · Editorial description reviewed October 2, 2026 |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Creative Commons Attribution 4.0 |