Poolside releases Laguna S 2.1, a 118B open-weight coding model with a 1M-token context

Poolside released Laguna S 2.1, an open-weight coding model with 118 billion total parameters, on July 21. The company pitches it as a mid-size option built for agentic software work, with a 1-million-token context window.

The model is a mixture-of-experts design that activates only about 8 billion parameters per token, which keeps inference costs closer to a small model while drawing on a much larger parameter pool. It slots between Poolside’s existing Laguna XS 2.1 and the larger Laguna M.1 in the company’s lineup.

The 1-million-token context, available in both “thinking” and “no-thinking” modes, targets long-horizon agentic coding tasks, where an assistant must track sprawling codebases and multi-step plans. Poolside said training took under nine weeks from the start of pre-training to launch.

Poolside reported benchmark scores of 70.2% on Terminal-Bench 2.1 with thinking enabled and 78.5% on SWE-Bench Multilingual. Those are the company’s own results and have not been independently verified.

Poolside released the weights under the OpenMDW-1.1 license in BF16, FP8, INT4 and NVFP4 precisions, with official GGUF and MLX conversions for local deployment. The model is also available through hosting providers including OpenRouter and Vercel AI Gateway.