Laguna XS.2 poolside laptop achieving 30.1% Terminal-Bench 2.0 score, outperforming Haiku 4.5 in performance comparison

Editorial illustration for Poolside's free Laguna XS.2 scores 30.1% on Terminal‑Bench 2.0, edging Haiku 4.5

Poolside's free Laguna XS.2 scores 30.1% on...

Poolside's free Laguna XS.2 scores 30.1% on Terminal‑Bench 2.0, edging Haiku 4.5

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

April 28, 2026 • 2 min read

Poolside’s latest release, Laguna XS.2, arrives as a free, open‑source option aimed at developers who want to run sophisticated agentic code locally without relying on cloud services. The startup frames the model as a bridge between heavyweight, proprietary systems and the emerging “nano” class of specialized AI, positioning it for hobbyists, startups, and research labs that need decent performance on modest hardware. By targeting terminal‑based reasoning tasks—a niche yet increasingly relevant benchmark for code generation and step‑by‑step problem solving—Poolside hopes to demonstrate that an openly available model can still hold its own against commercial contenders.

The numbers from Terminal‑Bench 2.0 will be the first concrete proof points for this claim, offering a side‑by‑side look at how XS.2 stacks up against both mainstream and ultra‑lean alternatives. Here’s the data that clarifies where Laguna XS.2 lands in that competitive mix.

As Poolside wrote in a blog post today , it's spent the last few years "focused on serving our government and public sector clients with capable models deployable into the highest-security environments," yet is now going open source "to support builders and the wider research community.

— Poolside, Poolside's free Laguna XS.2 scores 30.1% on Terminal‑Bench 2.0, edging Haiku 4.5 - VentureBeat AI

Is a free, open model enough to shift expectations? Laguna XS.2 shows promise, scoring 30.1 % on Terminal‑Bench 2.0—just enough to nudge ahead of Haiku 4.5’s 29.8 % and to demonstrate that high‑quality agentic coding can be offered without a price tag. Yet the gap to specialized “nano” models remains stark; GPT‑5.4 Nano registers 46.3 % on the same test, suggesting that XS.2’s advantage is limited to a narrow slice of the benchmark landscape.

The broader AI race, illustrated by Anthropic’s Claude Opus 4.7 and OpenAI’s GPT‑5.5, continues to unfold, and it is unclear whether a free offering can sustain relevance against rapidly evolving proprietary releases. Nonetheless, Poolside’s decision to make Laguna XS.2 publicly available adds a data point to the ongoing discussion about accessibility versus raw performance. Whether developers will adopt the model for local agentic tasks, or whether the modest edge over Haiku 4.5 will translate into practical gains, remains to be seen.

Poolside's free Laguna XS.2 scores 30.1% on...

Further Reading

Latest News