Poolside releases Laguna S 2.1, the West’s open rival to DeepSeek and Qwen

A San Francisco startup is making a bold play in the open-weight AI space, releasing a coding model it claims can go toe-to-toe with much larger rivals. Poolside has launched Laguna S 2.1, a 118-billion-parameter model designed for agentic coding, and is positioning it as the Western answer to the wave of Chinese open-weight models that have dominated the category.

The model employs a mixture-of-experts architecture, activating only eight billion parameters per token. This efficiency allows it to run on a single Nvidia DGX Spark desktop system. The weights are freely available on Hugging Face under the Linux Foundation’s OpenMDW license, giving enterprises and governments a self-hosted option.

On key benchmarks like Terminal-Bench and SWE-Bench Pro, Laguna S 2.1 scored just over 70% and nearly 60%, respectively. These results match or beat models from DeepSeek, Nvidia, and Thinking Machines that use two to eight times as many active parameters. Still, Poolside is candid about the model’s limitations, noting it is “not yet at the frontier.” Closed-source systems from OpenAI and Anthropic continue to score significantly higher.

This release is a direct response to a prolonged period of Chinese dominance in the open-weight category. For over a year, labs like DeepSeek, Alibaba’s Qwen family, and Moonshot’s Kimi have set the pace. According to Poolside, no Western lab had released an open-weight model in the 118-billion-parameter class for 11 months before this launch. Forbes reported that the company explicitly framed the release as a way to give Western enterprises and governments a self-hosted alternative that doesn’t require sending data to a foreign provider.

Founded in 2023 by Jason Warner, former CTO of GitHub, and Eiso Kant, Poolside raised $500 million in a Series B in October 2024, reaching a $3 billion valuation with backing from Nvidia and eBay. However, a planned $2 billion Series C that would have valued the company at $14 billion collapsed in April 2026 after CoreWeave withdrew from a joint data center project in Texas. Today, the company serves government, defense, and other highly regulated organizations through its API and agent harness.

Poolside built Laguna S 2.1 using its internal Model Factory platform, which automates architecture search and reinforcement learning from code execution. Training was completed in under four weeks on 4,000 Nvidia H200 GPUs. The smaller Laguna XS launched just three weeks earlier, and the company says it ships new models on a roughly five-week cadence. As a demonstration of long-horizon reasoning, Poolside published a trajectory of the model independently solving a combinatorics problem that only the largest frontier models had recently resolved.

The core bet is that enterprises will prefer to run a capable coding model on their own hardware rather than send prompts to a closed API. That thesis hinges on Laguna S 2.1 performing in production as well as it does on benchmarks. Poolside’s own results show the model trailing closed-source leaders by roughly 10 to 15 percentage points on Terminal-Bench, a gap that matters for customers weighing the trade-offs of self-hosting.

Whether Poolside can close that distance on its next cycle, while competing against Chinese open-weight models that are also improving rapidly, will determine if the Western open-weight gap is a temporary condition or a structural one.