Poolside Releases Open Source 118 Billion Parameter Model Laguna S 2.1, Self-Test Surpasses DeepSeek

Poolside has released the open-source MoE code model Laguna S 2.1, which has a total of 118 billion parameters, with 8 billion activated at a time, supporting a context of 1 million tokens. After quantization, it can run on a single NVIDIA DGX Spark. In self-testing, Laguna S 2.1 scored 70.2% on Terminal-Bench 2.1, surpassing DeepSeek-V4-Pro-Max, which has a total of 1.6 trillion parameters and 49 billion activated parameters, scoring 64.0%. It also performed better in SWE-Bench Multilingual, SWE-Bench Pro, and DeepSWE. However, Laguna S 2.1 still lags behind Kimi K3, which has a total of 2.8 trillion parameters and 50 billion activated parameters, scoring 88.3% on Terminal-Bench 2.1 and 69% on DeepSWE. Laguna S 2.1 scored 70.2% and 40.4%, respectively. The scores were not retested in the same environment; Poolside used its own Agent framework, while competitors took the highest public scores. The feature of Laguna S 2.1 is achieving long-range programming capabilities with fewer activated parameters for local single-machine operation.

Disclaimer: This content is provided for general branding and informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online events, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets or to use any services. Crypto assets are highly volatile and may result in loss. WEEX services and online events may not be available in all regions and are subject to applicable laws, regulations, and eligibility requirements. You are responsible for ensuring that your use of WEEX services complies with local laws and for carefully assessing the risks before participating in any crypto-related activities.