A compact multilingual generalist that punches modestly within its weight class. With only 560 million parameters, it handles text generation across dozens of languages — a notable trait inherited from BLOOM's large-scale multilingual training — but frequently struggles with coherence on complex or lengthy tasks. Think of it as a lightweight experiment-friendly model rather than a production workhorse.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| IFEval | 6.2 | accuracy | 1d ago |