LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
Laguna XS 2.1
◆ OPEN
by Poolside
OPEN WEIGHTSHOSTEDAPILONG CONTEXTPAIDFREE TO TRY
RANKED #209 OF 281 ASSISTANTS · OVERALL #332 OF 6,578 · VIBE SCORE 5.5 · 25 VOICES
Laguna XS 2.1 is the latest coding agent model in the 33B-A3B category from [Poolside](https://poolside.ai/) and a step forward from their Laguna XS.2 model (released in April 2026). It combines...
HAPPY10.0
6 mentions
SAD-
0 mentions
VIBE TRAJECTORY
26 weeks · 41 voices
AGING WELL
The crowd is warmer now than it was at the start.
HOW TO USE IT
Open weights. You can use a hosted service, or download it and run it yourself, free.
RUN IT ON YOUR OWN MACHINE
Open weights: download Laguna XS 2.1 once and it's yours. No subscription, no rate limits, works offline.
33B total parameters (3B active MoE), a Q4 GGUF is about 20 GB
- BARE MINIMUM
Nvidia GeForce RTX 3060 12GB plus 32GB system RAM
12 GB VRAM PLUS 32 GB RAM
~$220 USEDQ4_K_M partial GPU offload, ~12 tok/s, 8k context
- SWEET SPOT
Nvidia GeForce RTX 3090 24GB
24 GB VRAM
~$700 USEDQ4_K_M full GPU offload with FP8 KV cache, ~65 tok/s, 32k context
- FULL POWER
Apple Mac Studio M2 Ultra (64GB)
64 GB UNIFIED MEMORY
~$2,800 USEDFP8 or Q8 quantization, ~120 tok/s, 128k+ context
LEARN HOW TO SET IT UP
WILL MY BOX RUN LAGUNA XS 2.1?
The three cards above are researched picks. Search the machine you actually have — we only say yes if it should reply at a usable speed, not just load the weights and crawl.
HARDWARE PICKS & PRICES ARE RESEARCHED FROM THE LIVE WEB AND REFRESHED AUTOMATICALLY. TREAT THEM AS BALLPARK, NOT GOSPEL.
WHAT CAN YOU ACTUALLY DO WITH IT?
Everyday jobs on each of those rigs: how long Laguna XS 2.1 takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMNvidia GeForce RTX 3060 12GB plus 32GB system RAM | SWEET SPOTNvidia GeForce RTX 3090 24GB | FULL POWERApple Mac Studio M2 Ultra (64GB) |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~42 S | ~8 S | ~4 S |
| Summarize a documenta long report boiled down to the points that matter | ~2.1 MIN | ~23 S | ~13 S |
| Build a websitea small landing page, markup and styles together | ~11 MIN★★★★★ | ~2.1 MIN★★★★★ | ~67 S★★★★★ |
| Build a backendan API with routes, storage and tests | ~35 MIN★★★★★ | ~6.4 MIN★★★★★ | ~3.5 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~21 MIN★★★★★ | ~3.8 MIN★★★★★ | ~2.1 MIN★★★★★ |
STARS ARE THE COMMUNITY SCORE FOR THAT KIND OF WORK, DOCKED FOR HOW SQUASHED THE WEIGHTS GET AT EACH TIER. TIMES ARE COMPUTED FROM RESEARCHED SPEEDS. BALLPARK, NOT GOSPEL.
WHAT THE INTERNET SAYS
Not enough discussion about Laguna XS 2.1 yet to call it. We found 25 posts but people did not say much either way.
21 POSTS · 4 COMMENTS · DEV.TO · X · LEMMY · REDDIT · GITHUB · BLUESKY · BLOG
PEOPLE LOVE
- No fan mail yet. The crowd is still warming up.
PEOPLE COMPLAIN ABOUT
- No complaints yet. The internet is still looking for one. Suspicious.
LOOKING FOR AN ALTERNATIVE?
Other models the crowd has fully reviewed, starting with text models like this one.
DeepSeek V4 Flash> SOURCES (12)
- REDDIT · 2026-08-11 · 178 UPVOTESI gave DeepSeek V4 Flash basic vision by training a 40M connector on 100K examples
I gave DeepSeek V4 Flash basic vision by training a 40M connector on 100K examples
- X · 2026-08-15 · 42 UPVOTESWake up solvers! A new challenge just dropped! Last week the community made Laguna XS 2.1 run 2.
Wake up solvers! A new challenge just dropped! Last week the community made Laguna XS 2.1 run 2.
- BLOG · 2026-08-02 · 32 UPVOTESLatest open artifacts (#23): Laguna S2.1, Inkling, & Kimi K3 show the utility of open models on the Pareto frontier
Consolidation has been one of the paths that many astute observers predicted for the near-future of labs training models. It was labelled as inevitable, as training costs are increasing by orders of magnitude every year. Yet, as someone who in 2024 would’ve…
- X · 2026-08-15 · 20 UPVOTESOpen problems move faster in the open. Last week @yukonresearch ran on @poolsideai's Laguna XS 2
Open problems move faster in the open. Last week @yukonresearch ran on @poolsideai's Laguna XS 2
- REDDIT · 2026-08-17 · 7 UPVOTESBenchmarks iGPU integrated Radeon 680M models
Benchmarks iGPU integrated Radeon 680M models
- REDDIT · 2026-07-31 · 6 UPVOTESExperience sharing: How do you use your local models and for what kind of tasks?
Experience sharing: How do you use your local models and for what kind of tasks?
- REDDIT · 2026-08-08 · 5 UPVOTESIs deepseek v4 flash the best model as subagent?
Is deepseek v4 flash the best model as subagent?
- DEV.TO · 2026-08-01 · 5 UPVOTESLaguna XS 2.1 vs. Kimi K3: Engineering the Future of AI Coding Assistants
Laguna XS 2.1 vs. Kimi K3: Engineering the Future of AI Coding Assistants
- LEMMY · 2026-07-22 · 2 UPVOTESGlobal downloads of China's open-source AI models exceed 10 billion
Global downloads of China's open-source AI models exceed 10 billion
- GITHUB · 2026-08-22 · 0 UPVOTESfix: parse specialized llama.cpp chat templates
fix: parse specialized llama.cpp chat templates
- REDDIT · 2026-08-16 · 0 UPVOTESAre ~50B models are good enough for Great Coding? With 32GB VRAM + 128GB RAM
Are ~50B models are good enough for Great Coding? With 32GB VRAM + 128GB RAM
- GITHUB · 2026-08-15 · 0 UPVOTESPer-model effective context: configured -c 131072 vs qwen3-1.7b's live n_ctx 40960
Per-model effective context: configured -c 131072 vs qwen3-1.7b's live n_ctx 40960