Laguna XS 2.1: benchmark results
Poolside's open-weight Laguna XS 2.1 (July 2026), a 33-billion-parameter, 3-billion-active MoE model for agentic coding and long-horizon work on local machines, upgraded from Laguna XS.2. Provider: Poolside. Released 2026-07-02. Access: Open.
Aggregate Rating 1590 ± 22, rank #485 of 1635 rated models, from 25 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM Stats Score | 24.24 | LLM Stats Score (conservative rating) | 51.6 |
| AI Chess Leaderboard (Reasoning) | 711 | Elo | 51 |
| ObviousBench | 92.36 | Answer pass³ (%) | 50.4 |
| Enkrypt AI - Toxicity Risk | 3.64 | Risk Score | 47.3 |
| AI Chess Leaderboard (Continuation) | 579 | Elo | 44.9 |
| ReasonScape R12 | 778.71 | ReasonScore | 40.2 |
| SpeechMap Compliance | 53.7 | % Requests Completed | 37.7 |
| SvelteBench | 81.1 | Average pass@1 (%) | 31.4 |
| Enkrypt AI - Insecure Code Risk | 41.33 | Risk Score | 31.1 |
| LLM Stats (SWE-bench Multilingual) | 63.1 | Score (%) | 28.6 |
| Enkrypt AI - Bias Risk | 87.34 | Risk Score | 24.7 |
| Tinybird AI SQL Benchmark - Exactness | 40.56 | Result exactness vs human reference queries (0-100) | 21 |
Interactive version: theaggregate.ai/model?slug=laguna-xs-2-1 · How It Works · Data refreshed daily, snapshot 2026-11-05.