Poolside Laguna XS 2.1 is the small, fast tier of Poolside's coding-native line and the most recent model in the registry. On OpenRouter it costs $0.06/$0.12 per Mtok while holding the same 262,144 token context as its bigger sibling, with 32,768 max output. Poolside's first-party API lists it at $0.00/$0.00 per Mtok, and OpenRouter carries a free variant too. Inside Atlas, Laguna XS 2.1 is the `small_model` that keeps your cheap slot inside a coding-native family instead of handing overhead work to a general-purpose model.
How does Poolside Laguna XS 2.1 compare to Laguna M.1 on price?
Poolside Laguna XS 2.1 carries the same 262,144 token context as Laguna M.1 at roughly a third of the paid price, $0.06/$0.12 per Mtok versus $0.20/$0.40 per Mtok on OpenRouter. The window is identical. The capability tier is not.
Laguna XS 2.1 is the fast tier and Laguna M.1 is the reasoning tier, but Poolside gives both the same quarter-million-token window. That makes XS 2.1 unusually capable as a cheap slot: most `small_model` candidates force you to accept a small context along with the small price, and Laguna XS 2.1 does not. Output on Laguna XS 2.1 is only 2x input, so at $0.12 per Mtok out it is one of the cheapest reasoning-capable models with a quarter-million-token window. Inside Atlas, that means summarization and subagent work can see a lot of code without costing much.
How do you pair Laguna XS 2.1 and Laguna M.1 in one Atlas config?
In Atlas, set `"small_model": "openrouter/poolside/laguna-xs-2.1"` and `"model": "openrouter/poolside/laguna-m.1"` so both slots stay inside a coding-native family. Laguna XS 2.1 runs $0.06/$0.12 per Mtok and Laguna M.1 runs $0.20/$0.40 per Mtok.
Keeping both Atlas slots inside Poolside's line is a coherent choice rather than a cost hack. Poolside trains exclusively for software engineering, so the model summarizing your session and the model writing your diff share the same training objective. Add both to favorites in the `/models` dialog so `model.cycle_recent` flips between the fast and the reasoning tier without opening a dialog. Atlas lets you switch the active model and provider on the fly with favorites and recents, so a task that turns out harder than expected can move from Laguna XS 2.1 to Laguna M.1 mid-session without restarting the work.
Can you try Poolside Laguna XS 2.1 for free?
Yes. OpenRouter carries a `poolside/laguna-xs-2.1:free` variant at $0/$0 per Mtok, so you can measure Laguna XS 2.1 against your repo before committing. Poolside's own first-party API also lists it at $0.00/$0.00 per Mtok.
Two free routes exist for Laguna XS 2.1: the `poolside/laguna-xs-2.1:free` variant on OpenRouter, and Poolside's first-party API at $0.00/$0.00 per Mtok. Either is enough to evaluate the model against a real codebase before spending. Atlas makes that evaluation meaningful because of what it feeds the model. Atlas searches code with hybrid semantic and keyword retrieval fused by reciprocal rank fusion, and Atlas indexes code by AST declarations using tree-sitter, not blind line windows, so Laguna XS 2.1 is judged on real declarations from your repo rather than on a generic benchmark. If it holds up, move to the paid $0.06/$0.12 per Mtok OpenRouter listing for stability.
What are the limits of Poolside Laguna XS 2.1?
XS means extra small. Poolside Laguna XS 2.1 is the fast tier, and it will not match Laguna M.1 or a 120B MoE on a refactor that spans many files. Max output is 32,768 tokens against a 262,144 context, so it reads far more than it can write in one turn.
Laguna XS 2.1's two limits are worth stating before you promote it to the build slot. Capacity is the first: a fast tier model is a fast tier model, and a multi-file refactor with cross-file invariants is exactly where the gap to Laguna M.1 shows. Output ceiling is the second: 32,768 max output against 262,144 tokens of context means the read-to-write ratio is roughly eight to one. Inside Atlas, that ratio is manageable because Atlas computes a unified diff for every file edit and surfaces it for approval before writing, and Atlas snapshots file changes as git patches so edits can be diffed and rolled back.
How do you set up Atlas with Poolside Laguna XS 2.1?
Setting up Poolside Laguna XS 2.1 in Atlas takes three steps: export POOLSIDE_API_KEY for the first-party route or OPENROUTER_API_KEY for the $0.06/$0.12 paid route, run `atlas models poolside`, and set the `small_model` key in atlas.json.
Poolside Laguna XS 2.1 resolves through either provider, so pick based on whether you want the free capacity-limited path or the paid OpenRouter listing. Run `atlas models poolside` (or `atlas models openrouter`) and find the laguna-xs-2.1 row before editing atlas.json, so a mistyped model id surfaces as a missing row rather than a runtime error. Then set `"small_model": "openrouter/poolside/laguna-xs-2.1"` and `"model": "openrouter/poolside/laguna-m.1"`. Every Atlas tool call is permission-gated against allow, ask, and deny rules before it runs, so testing a new model on a live repository is bounded from the start.
Setup
- 01Export POOLSIDE_API_KEY for the first-party route, or OPENROUTER_API_KEY for the $0.06/$0.12 paid route.
- 02Run `atlas models poolside` (or `atlas models openrouter`) and find the laguna-xs-2.1 row.
- 03Set `"small_model": "openrouter/poolside/laguna-xs-2.1"` and `"model": "openrouter/poolside/laguna-m.1"` so both slots stay inside a coding-native family.
- 04Add both to favorites in the `/models` dialog so `model.cycle_recent` flips between the fast and the reasoning tier without opening a dialog.
- 05Measure Laguna XS 2.1 against your repo first on the `poolside/laguna-xs-2.1:free` variant at $0/$0 per Mtok.
Frequently asked questions
- how much does poolside laguna xs 2.1 cost
- OpenRouter lists Poolside Laguna XS 2.1 at $0.06/$0.12 per Mtok, with a free tier at $0.00/$0.00. Poolside's first-party API also lists it at $0.00/$0.00 per Mtok.
- laguna xs 2.1 vs laguna m.1
- Laguna XS 2.1 is the fast tier at $0.06/$0.12 per Mtok. Laguna M.1 is the reasoning tier at $0.20/$0.40 per Mtok on OpenRouter. Both hold 262,144 tokens of context, but XS will not match M.1 on a refactor that spans many files.
- what is the context window of laguna xs 2.1
- Poolside Laguna XS 2.1 holds 262,144 tokens of context with 32,768 max output, the same window as its larger sibling Laguna M.1.
- how do I set laguna xs 2.1 as atlas small_model
- Export OPENROUTER_API_KEY, run `atlas models openrouter` and find the laguna-xs-2.1 row, then set `"small_model": "openrouter/poolside/laguna-xs-2.1"` in atlas.json alongside `"model": "openrouter/poolside/laguna-m.1"`.
- is there a free version of poolside laguna xs 2.1
- Yes. OpenRouter carries a `poolside/laguna-xs-2.1:free` variant at $0/$0 per Mtok, and Poolside's first-party API lists Laguna XS 2.1 at $0.00/$0.00 per Mtok.
- can laguna xs 2.1 be my main coding model
- Probably not. XS means extra small. Laguna XS 2.1 is the fast tier, and it will not match Laguna M.1 or a 120B MoE on a refactor that spans many files. Use it in the `small_model` slot.
- cheapest coding native model with a large context
- Poolside Laguna XS 2.1 at $0.06/$0.12 per Mtok on OpenRouter holds 262,144 tokens. Output is only 2x input, making it one of the cheapest reasoning-capable models with a quarter-million-token window.
Try Atlas in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install AtlasRelated guides
Atlas vs Graphite: Terminal AI Coding Agents in 2026
Comparing Atlas and Graphite in 2026: Atlas offers a terminal-native AI coding agent with local privacy, while Graphite focuses on stacked PR workflows and GitHub integration.
Atlas vs JetBrains Junie in 2026: Terminal Agent vs IDE Debugger-Driving Agent
Atlas vs JetBrains Junie in 2026: Junie drives the IDE debugger and reports 61.6% resolved on SWE-Rebench, bundled from $10/mo. Atlas is a free, open source terminal agent with permission gates.
Atlas vs Sourcegraph Cody: Terminal AI Coding Agents in 2026
Comparing Atlas, the terminal-native AI coding agent, with Sourcegraph Cody, an editor extension for whole-codebase search, for developers in 2026. Explore features, pricing, and workflow.
Atlas for Fastify in 2026
Atlas is a terminal-native AI coding agent for Fastify in 2026. It reads plugin encapsulation and JSON schemas, then runs node --test behind a permission prompt.
Run Atlas Headless in CI with Atlas (2026 Workflow)
How to run Atlas headless in CI in 2026: atlas run sends one prompt and exits when the session goes idle, with --format json, --command, and --continue for pipeline steps.
Atlas for Rust in 2026
Adopt Atlas, the terminal-native AI coding agent, for Rust development in 2026. Tackle borrow checker errors and clippy lints with Atlas's secure, approval-gated assistance.
Atlas vs Codebuff: Terminal AI Coding Agents in 2026
Atlas and Codebuff are terminal AI coding agents for 2026. Compare Atlas's terminal-native TUI, permission-gated tools, and diff review with Codebuff's multi-agent system and flexible pricing.
Atlas for Pandas: Terminal-Native AI Coding in 2026
Atlas is a terminal-native AI coding agent for Pandas. Vectorize df.apply, fix chained assignment under Copy-on-Write, and pin DataFrames with assert_frame_equal.