Across 6 shared benchmarks, Claude Fable 5 scores higher on 6 and laguna-s-2.1 on 0. The widest gap is DeepSWE, where Claude Fable 5 scores 70 against 40.4. laguna-s-2.1 is the cheaper of the two on tracked API pricing ($0.09 against $10.00 per million input tokens).
| Benchmark | Claude Fable 5 | laguna-s-2.1 |
|---|---|---|
| DeepSWE | 70 | 40.4 |
| DeepSWE 1.1 | 70 | 40.4 |
| SWE-bench Multilingual | 86.6 | 78.5 |
| SWE-bench Pro | 80 | 59.4 |
| Terminal-Bench 2.1 | 88 | 70.2 |
| toolathlon | 61.7 | 49.7 |
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, poolside), otherwise the lowest tracked offer.