Laguna S 2.1: benchmark results
Poolside's open-weight Laguna S 2.1 (July 2026), a 118-billion-parameter, 8-billion-active MoE model for agentic coding and long-horizon work, sitting between Laguna XS 2.1 and Laguna M.1 in the family. Provider: Poolside. Released 2026-07-21. Access: Open.
Aggregate Rating 1594 ± 22, rank #462 of 1656 rated models, from 39 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM Stats (SWE-bench Multilingual) | 78.5 | Score (%) | 85.1 |
| LLM Stats Score | 39.48 | LLM Stats Score (conservative rating) | 82.1 |
| AI Chess Leaderboard (Reasoning) | 1183 | Elo | 81.6 |
| AI Chess Leaderboard (Continuation) | 962 | Elo | 77.7 |
| Tinybird AI SQL Benchmark - Success Rate | 100 | Questions answered with a valid query within 3 retries (%) | 71.7 |
| Enkrypt AI - Toxicity Risk | 2.73 | Risk Score | 58.7 |
| ALEM (Multi-Agent Coordination) | 7.4 | Total Achievement % (medium) | 55.9 |
| Enkrypt AI - Insecure Code Risk | 25.33 | Risk Score | 54.4 |
| Tinybird AI SQL Benchmark - Exactness | 47.69 | Result exactness vs human reference queries (0-100) | 44.2 |
| Bullshit Benchmark | 23.6 | BS Detection Rate (%) | 41.5 |
| LLM Stats (Terminal-Bench 2.1) | 70.2 | Score (%) | 37.8 |
| Tinybird AI SQL Benchmark - First-Attempt Success Rate | 90 | Questions answered with a valid query on the first attempt ( | 34.8 |
Interactive version: theaggregate.ai/model?slug=laguna-s-2-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.