San Francisco AI startup Poolside has launched Laguna S 2.1, a 118-billion-parameter open-weight coding model designed to rival much larger competitors while fitting on a desktop. The release is positioned as a strategic Western alternative to dominant Chinese coding AI models.
- 118B parameter open-weight model with mixture-of-experts architecture
- Runs on single Nvidia DGX Spark desktop hardware
- Aims to counter Chinese dominance with Western self-hosted alternative
What happened
Poolside has introduced Laguna S 2.1, a large-scale open-weight coding model with 118 billion parameters. The model uses a mixture-of-experts approach enabling eight billion active parameters per token, delivering high efficiency for agentic coding tasks. It is compact enough to run on a desktop Nvidia DGX Spark system, enabling self-hosted deployment.
The model was trained using Poolside’s internal Model Factory platform in under four weeks on 4,000 Nvidia H200 GPUs. It has demonstrated competitive results against larger models from DeepSeek, Nvidia, and Thinking Machines on benchmarks such as Terminal-Bench and SWE-Bench Pro. The release marks Poolside’s first open-weight model at this scale in nearly a year.
Why it matters
Laguna S 2.1 is positioned as a response to the prominence of Chinese open-weight coding models like DeepSeek, Alibaba’s Qwen family, and Moonshot’s Kimi, which have led the category for over a year. No Western lab had released an open-weight model at this scale since then, highlighting a gap Poolside seeks to close.
The availability of an open-weight alternative is crucial for Western enterprises and governments prioritizing data privacy and control. Poolside’s model offers a large-scale coding AI that can be deployed on-premises, providing an option that does not require sending sensitive data to foreign cloud providers or closed APIs.
What to watch next
While Laguna S 2.1 performs well, it still lags by about 10 to 15 percentage points behind the leading closed-source models from OpenAI and Anthropic according to key benchmarks. Whether Poolside can close this gap in upcoming iterations will be critical to gaining broader enterprise adoption.
Competition will also intensify as Chinese open-weight models continue improving rapidly. Poolside’s ability to innovate on performance, scalability, and deployment ease will determine if the Western open-weight model market is a transient catch-up effort or the start of a sustainable competitive presence.