Sulat.com
AI models
$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3 from Synthetic
Kilo Gateway logo

Model details

Poolside: Laguna XS 2.1 (retires Oct 31)

Poolside's Laguna XS 2.1 is positioned as the latest coding agent model in the "33B-A3B" category from the company's Laguna family, succeeding the Laguna XS.2 release from April 2026 with a step forward aimed at agentic software development workflows. The architecture framing suggests a 33B-parameter configuration with an A3B-style design, aligning it with the emerging class of mid-sized models that pair broad language modeling with stronger active routing for structured, tool-using tasks. Poolside's broader focus on applied AI for software engineering carries into this release, with the model surfaced through NVIDIA's NIM integration under the poolside/laguna-xs-2.1 identifier and reachable via an OpenAI-compatible chat completions endpoint that exposes temperature and top-p sampling controls for response calibration.

In practical terms, Laguna XS 2.1 is tuned for code generation and longer-horizon agent tasks that benefit from very large context spans, making it a reasonable fit for repository-scale reasoning, multi-file refactors, and IDE-integrated development agents where sustained instruction following over long inputs is valuable. Its availability through Kilo's coding agent environment, alongside NVIDIA NIM and pricing routed via OpenRouter, gives developer-facing tooling multiple entry points for evaluation and production use. The combination of an extended context window, agent-oriented design, and accessible sampling controls makes it well suited to teams exploring coding assistants that need to coordinate tool calls and temperature-managed generation within a single, long-running session.

Kilo Gatewaypoolside/laguna-xs-2.1laguna

Quick Info

Powered by
Provider
Kilo Gateway
Model key
poolside/laguna-xs-2.1
Release date
Jul 2, 2026
Last updated
Jul 2, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$0.10
Output token cost
$0.20

Limits

Output tokens
32,768 tokens
Context window
262,144 tokens

Transparent token rates

Compare Poolside: Laguna XS 2.1 (retires Oct 31) pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Poolside: Laguna XS 2.1 (retires Oct 31)

Kilo Gateway

CoverageRelease Notes

Poolside announced Laguna XS 2.1 on July 2, 2026, updating its open-weight coding model line with a 33-billion-parameter Mixture-of-Experts design that activates roughly 3 billion parameters per token. Per the report, the architecture uses 40 layers and 256 experts and ships with a 262,144-token context window in the m The release emphasizes a measurable coding-benchmark lift over the prior XS.2, most notably a 5.4-point gain on SWE-bench Multilingual to 63.1%. Poolside and Hugging Face cite FP8 KV-cache quantization, day-one support in vLLM and TRT-LLM, and multiple quantized checkpoints (FP8, INT4, NVFP4) to reduce VRAM and acceler

Kilo Gateway

Coverage

A third-party test dated July 3, 2026 ran Laguna XS 2.1 locally across nine programming languages, presenting it as a 33B-total / 3B-active MoE compact coding model positioned as a local alternative to cloud coding APIs. The report frames the model as software-engineering-focused rather than general-purpose, and notes The piece claims Laguna XS 2.1 scores above both North Mini Code and Claude Haiku 4.5 on SWE-bench Pro, a benchmark built from real GitHub issues, arguing this matters because it holds on practical issue-resolution tasks while staying deployable on local hardware. The content is largely a write-up of a YouTube walkthro

Videos about Poolside: Laguna XS 2.1 (retires Oct 31)

More models around Poolside: Laguna XS 2.1 (retires Oct 31)