Laguna S 2.1

Laguna S 2.1 lists base rates of $0.09 per million input tokens and $0.18 per million output tokens. Estimate both input and output token volume and any additional charges for your workload. Poolside route on OpenRouter $0.09/$0.18 per MTok, cache read $0.009, FP4 serving, 1,048,576 context and 131,072 maximum completion. The feed reports a 10% discount; no end date is stated. These rates and output cap are route-specific. Direct Poolside inference documents max_tokens default 32,768 and accepted range 1–262,144; this is a request range, not the OpenRouter cap. Downloaded weights use OpenMDW-1.1 and incur compute costs.

Laguna S 2.1 accepts about 1.05M tokens of context. That only matters if you routinely send very long documents, large codebases, or multi-turn histories that approach that limit.

Laguna S 2.1 is a balanced tier from Poolside. It suits long-context coding, agentic research, self-hosting. Thinking off/max are the two model behaviors, max enabled by default; direct API toggles enable_thinking, not a low/medium/high effort enum. Preserve prior reasoning_content. Native context is 1,048,576. OpenMDW-1.1 redistribution requires retaining the agreement and applicable notices. July 21 vendor scores use pool harness with four attempts per task for Terminal and three for DeepSWE; the comparison table mixes maximum scores across different harnesses, so cross-model verdicts are withheld.

Laguna S 2.1 publishes open weights, but self-hosting depends on the licence, hardware footprint, quantisation quality, and serving stack. A hosted API is often cheaper until you have measured throughput and concurrency on your own hardware.