Stable LM Zephyr 3B
Stable LM Zephyr 3B is an English instruction model adapted from stablelm-3b-4e1t. Its training uses supervised dialogue and instruction data followed by Direct Preference Optimization.
Model design and access
The card identifies instruction sources such as UltraChat, MetaMathQA and SlimOrca, plus preference datasets. Its published license differs from the non-commercial terms on Stable LM 2 Zephyr 1.6B. Similar Zephyr names therefore do not establish matching usage rights.Stable LM Zephyr 3B official model card
Practical use
Practical guidance. Zephyr 3B provides a compact local dialogue baseline with a distinct licensing position. Compare it using the application's own prompts and output requirements. Historical benchmark scores can explain the release, but they should not be treated as current evidence that it outperforms later models.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | stablelm-zephyr-3b |
| Context window | 4,096 tokens Published checkpoint sequence length; prompt and generated text share the window. Stable LM Zephyr 3B configuration |
| Base model | stablelm-3b-4e1t.Stable LM Zephyr 3B official model card |
| License | Stability AI Community License.Stable LM Zephyr 3B official model card |
| Release position | Earlier 3B family, separate from Stable LM 2.Stable LM Zephyr 3B official model card |
Related models
- Stable LM 2 Chat 1.6B
Stability AI · Stable LM
Stable LM 2 Chat 1.6B is an English instruction-tuned version of the 1.6B base.
Launch date not confirmedResearch releaseDownloadable - Stable LM 2 Zephyr 1.6B
Stability AI · Stable LM
Stable LM 2 Zephyr 1.6B is an earlier instruction-tuned release based on Stable LM 2 1.6B.
Launch date not confirmedResearch releaseDownloadable