Muse Glimmer 30B open weights tested
Fifteenth on the extended CymraegBench leaderboard with an overall score of 70.64.
The facts
| Developer | Meta |
|---|---|
| OpenRouter ID | meta/muse-glimmer-30b |
| Size | 30 billion parameters |
| Test pricing | $0.30 / $1.20 per million tokens (input/output) |
| Context | 131,072 tokens; up to 16,384 output tokens |
| Full run cost | $0.62 |
Model source: Muse Glimmer 30B’s Hugging Face card.
CymraegBench result
| Overall score | Welsh reasoning | Practical Welsh | Translation BLEU |
|---|---|---|---|
| 70.64 | 77.50 | 63.78 | 60.71 |
The same sample was used as for the other leading models: up to 100 cases from each of the 19 core suites, with reasoning_effort=low and an output limit of 8,192 tokens. The 232 CCC cases do not contribute to the overall score.
Glimmer is fifteenth on the extended leaderboard: 2.74 points behind GLM-5.3 and 0.17 points ahead of Qwen3.8 27B. It completed all 2,053 cases without a refusal or invalid output.
CCC Language Audit
| Subset | Score |
|---|---|
| Correction and explanation | 61.72 |
| Idiom | 96.88 |
| Terminology | 87.50 |
| Summarisation and quotation | 91.07 |
| References | 90.63 |
| Welsh prompt | 100.00 |
| English prompt | 85.83 |
| Average across the seven subsets | 87.66 |
Glimmer scored 65.63 for correction and 35.94 for explanation. On the references test it scored 100 for selecting references, 68.75 for answer content, and 100 for not inventing a source.
See the full leaderboard, CCC Language Audit, or download the summary, scores by suite, CCC dimensions and the metadata.