The Sequence AI of the Week #903: Laguna, the 118 Billion Parameters that Walks Into a Trillion-Parameter Bar
Poolside’s 118B coding model beats systems ten times its size. The interesting part is not the architecture.
TL;DR
- Open-weight models generally show a positive correlation between parameter count and benchmark scores.
- Laguna S 2.1 (118B parameters) scores 70.2% on Terminal-Bench 2.1, outperforming larger models like DeepSeek-V4-Pro-Max (1.6T parameters, 64.0%) and Inkling (975B parameters, 63.8%).
- On the DeepSWE benchmark, Laguna S 2.1 (40.4%) significantly outperforms DeepSeek-V4-Pro-Max (9.0%), despite a 13x parameter deficit.
- Poolside has published evaluation trajectories to validate the results, indicating transparency in the release.