Laguna S 2.1
poolside · Released Jul 2026
Laguna S 2.1 is the first credible Western open-weight coding model in 11 months — a 118B MoE that activates only 8B parameters per token yet scores 70.2% on Terminal-Bench 2.1, beating models 10x its size. But it's not at the frontier — GPT-5.6 Sol and Kimi K3 sit far ahead — and known limitations include harness overfitting, JSON mangling in nested tool arguments, and no configurable thinking-effort control. Poolside's collapsed $14B Series C raises sustainability questions, making this best for self-hosted agentic coding where data sovereignty matters more than absolute benchmark leadership.
Is it right for you?
Good for
- Agentic coding on open weights: Poolside reports 70.2% on Terminal-Bench 2.1 from a 118B mixture-of-experts model with 8B active parameters per token (vendor-reported, untested by us)
- Self-hosting or shopping between hosts: weights are on Hugging Face under OpenMDW-1.1 and the model is served on Baseten, OpenRouter and Vercel AI Gateway, with a 1M-token context on paid endpoints
Not good for
- Harness portability: Poolside's own launch post lists harness overfitting, where slight differences in tool schemas across agent frameworks hurt results
- Nested tool arguments: Poolside lists trouble with nested tool calls involving JSON arrays as a known limitation
- Hard math and long reasoning: Poolside lists longer than expected thinking duration, especially on complex mathematics problems
Pricing
Input
$0.10 / 1M
Output
$0.20 / 1M
Context
OpenMDW-1.1 license, 1M ctx (256K on the OpenRouter free endpoint). $0.10 in / $0.20 out / $0.01 cache read is OpenRouter's paid endpoint rate per Poolside's launch post; Poolside lists the models as free to use for a limited time
Benchmarks
No verdict changes yet
The clock starts day one. Changes land here as our verdict evolves.
Sources
Verification log
- Pricing· Updated
Automated agent
Price figures unchanged ($0.10 in / $0.20 out). Previously unsourced: Poolside's launch post attributes them to OpenRouter's paid endpoint (cache read $0.01, 1M ctx; free endpoint 256K), while poolside.ai/models says the models are free to use for a limited time. Rewrote pricing_context accordingly, added sourced strengths, limitations and benchmarks (Terminal-Bench 2.1 70.2%, DeepSWE v1.1 40.4%, Poolside-reported). Stored summary carries an unsourced Series C claim and an em dash; correction staged as entity_verdict.
- Pricing· No changes
Automated agent
- Pricing· Needs attention
Automated agent
Stored $0.10 / $0.20 per 1M has no source on the page; the page prints no price and says the API is free for a limited time. Stored value kept unchanged (not overwritten with a guess). Entity still has empty sources, strengths, audiences and benchmarks and a summary with an em dash; needs a sourced rebuild.
- Pricing· Needs attention
Automated agent
Poolside publishes no per-token price and says free to use for a limited time; the blog URL returns 404 and OpenRouter shows no price. Stored $0.10/$0.20 unsourced and kept.
- Pricing· Needs attention
Automated agent
Page lists 118B/8B-active, 1M context, open weights and free limited-time evaluation access but no per-token price, so the stored $0.10/$0.20 could not be confirmed or refuted. Stored values kept.
- Pricing· Needs attention
Automated agent
- Pricing· Needs attention
Automated agent
Poolside publishes no per-token price; the stored $0.10/$0.20 has no vendor source and the entity has no sources children. Stored value kept, not overwritten or blanked.
- Pricing· Needs attention
Automated agent
Poolside's models page no longer lists a per-token price for Laguna S 2.1. It says 'Licenses vary by model' and mentions '1M context' without tying it to this model. The OpenRouter page and the models API returned no Laguna data. The stored $0.10/$0.20 and OpenMDW-1.1 license are kept as unverified.
- Pricing· No changes
Automated agent
- Pricing· No changes
Automated agent