Across 16 shared benchmarks, Claude Opus 4.8 scores higher on 13 and muse-glimmer-30b on 3. The widest gap is HLE, where Claude Opus 4.8 scores 57.9 against 22. muse-glimmer-30b is the cheaper of the two on tracked API pricing ($0.30 against $5.00 per million input tokens).
| Benchmark | Claude Opus 4.8 | muse-glimmer-30b |
|---|---|---|
| AA-LCR | 73 | 80 |
| aime_2026 | 100 | 94.7 |
| charxiv_rq | 80.5 | 85.9 |
| deepsearchqa_f1 | 93.1 | 74.6 |
| GPQA Diamond | 93.6 | 83.5 |
| HLE | 57.9 | 22 |
| IFBench | 62.2 | 77 |
| MCP Atlas | 83.6 | 75.5 |
| MMMU-Pro | 78.9 | 74 |
| OSWorld-Verified | 83.4 | 65.9 |
| scicode | 53.5 | 43.6 |
| screenspot_pro_no_tools | 87.9 | 75.4 |
| SWE-bench Pro | 69.2 | 51.2 |
| SWE-bench Verified | 88.6 | 76 |
| Tau 3 Banking | 27.6 | 23.5 |
| Terminal-Bench 2.1 | 85 | 51.7 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing. Quoted rates are the price-setter row we currently track for each model — its direct or vendor-official listing where one exists (direct, deepinfra), otherwise the lowest tracked offer.