Laguna S 2.1
Poolside · 2026-07-21 · 117.6B parameters
Laguna S 2.1 is Poolside's open-weight 117.6B-total, approximately 8B-active Mixture-of-Experts model for agentic coding and long-horizon work. It supports optional interleaved thinking and uses 12 full-attention layers (48 query heads, YaRN-scaled RoPE) interleaved with 36 512-token sliding-window layers (72 query heads, standard RoPE), across a 1,048,576-token context.
Benchmark scores
| Benchmark | Score |
|---|---|
| BuseyBench SVG | 2.9 |
| Capability | 155.1 |
| DeepSWE v1.1 | 40.4 |
| SWE Atlas - Codebase QnA | 46.2 |
| SWE-bench Multilingual | 78.5 |
| SWE-Bench Pro (Public Dataset) | 59.4 |
| Terminal-Bench 2.1 | 70.2 |
| Toolathlon-Verified | 49.7 |