Laguna S 2.1 Free

OpenCode Zen's official documentation describes Zen as a curated model gateway that vets and benchmarks coding models, then serves them via the https://opencode.ai/zen/v1 endpoint with per-request billing and AI SDK packages. Integration is handled through the /connect and /models commands in the OpenCode TUI after obt The supplied excerpt only enumerates GPT 5.x family endpoints and is truncated before any Laguna-family entries, so it does not explicitly confirm Laguna S 2.1 Free's presence on Zen. It still provides the authoritative API surface and integration flow for any Zen-hosted model, including the subject, and is useful as t

Poolside released Laguna S 2.1 on July 22, 2026 as an open-weight coding model built on a Mixture-of-Experts architecture with 118 billion total parameters but only 8 billion active per token. The model offers a 1-million-token context window and runs in separate thinking and no-thinking modes, with a training run that OpenCode added support for the model on the same day, exposing the free tier with the full 1M-token context through its agent harness with no separate API key or hosting setup required. Poolside co-CEO Eiso Kant framed the release around efficiency rather than raw parameter count, positioning Laguna S 2.1 as persistent

WebBrain benchmarked Poolside's Laguna S 2.1 (118B total / 8B active MoE, 48 layers, native 1M context) on its frozen 100-case browser-agent first-tool benchmark via OpenRouter. A default run landed at 65% Sonnet alignment, and an explicit high-reasoning request pushed it to 71%, still trailing Tencent Hy3 (73%) and Mi The article contextualizes Poolside's published benchmark table: Terminal-Bench 2.1 at 70.2 (vs Hy3 71.7), SWE-Bench Multilingual 78.5 (vs Hy3 75.8), SWE-Bench Pro public 59.4 (vs Hy3 57.9), DeepSWE 40.4, SWE Atlas 46.2, and Toolathlon Verified 49.7. It flags a methodology caveat that pass@1 is averaged over four attem

Kilo Code's catalog page for the free variant lists the model id as poolside/laguna-s-2.1:free, created July 21, 2026, with a 262,144-token context window, a 32,768-token maximum output, text-only input modality, and $0.00 per million tokens for both input and output. The page describes Laguna S 2.1 as Poolside's lates Kilo's leaderboard ranks the model at 46 in Code mode, 90 in Ask, 78 in Debug, and 64 in Orchestrator for the prior week, with usage charts available for late September 2026. The page marks the model as recommended for Code mode and currently available at no cost. Note that the 262,144-token context window reported her

OpenRouter's page for the exact variant `poolside/laguna-s-2.1:free` identifies it as Poolside's latest coding agent model: a 118B total / 8B active parameter model scoring 70.2% on Terminal-Bench 2.1 and 40.4% on DeepSWE, designed for software engineering and agentic coding use cases. The page states the model is open The listing reports a context window of 262K tokens, a release date of July 21, 2026, and pricing of Free with a single upstream provider (Poolside Free) at 35 tok/s P50 throughput and 1.85s P50 latency, with reported uptime of 100.00%. It also surfaces a measured tool-call error rate of about 9.77% and a cache hit rat

Laguna S 2.1 from Poolside became available on Vercel AI Gateway with an explicit free/paid id split: `poolside/laguna-s-2.1-free` for the free tier and `poolside/laguna-s-2.1` for higher throughput and rate limits. The Tech Bytes article (referencing the Vercel blog) instructs users to point AI Gateway requests at the The model is described as an open-weight Mixture-of-Experts release with a context window of up to 1 million tokens and support for both thinking and no-thinking modes. Operators choosing the Free variant should confirm the 1M-token context applies to their selected id and that the thinking/no-thinking behavior matches

LM Market Cap's AI Market Outlook 2026 lists "Laguna S 2.1 (free)" from poolside as a recent coding-model launch approximately 6 days before July 27, 2026, placed alongside the paid "Laguna S 2.1" entry in the "Recent launches" section. This is the closest supplied evidence to the subject, indicating that a free tier o The page is a third-party analytics dashboard, not an official OpenCode Zen or poolside announcement, so it does not provide context window, pricing, benchmark, or API details for the free tier. Provider attribution to poolside is aggregator-side, and it is unclear whether "Laguna S 2.1 (free)" is a separate SKU on Ope