LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
Laguna S 2.1
◆ OPEN
by OpenAI
OPEN WEIGHTSLOCALAPILONG CONTEXTPAIDFREE TO TRY
RANKED #156 OF 231 ASSISTANTS · OVERALL #248 OF 6,583 · VIBE SCORE 5.6 · 94 VOICES
Laguna S 2.1 is the latest coding agent model from [Poolside](<https://poolside.ai/>). Laguna S 2.1 is a 118B total parameter model with 8B active parameters, scoring 70.2% on Terminal-Bench 2.1 and...
HAPPY8.3
13 mentions
SAD6.7
7 mentions
VIBE TRAJECTORY
26 weeks · 145 voices
QUIETLY DEGRADING
Recent vibe and chatter are both sliding.
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 S 2.1 once and it's yours. No subscription, no rate limits, works offline.
118B parameters (8B active MoE), a Q4 file is about 75 GB
- BARE MINIMUM
Apple Mac Studio M2 Ultra (128GB Unified Memory)
128 GB UNIFIED MEMORY
~$3500 USEDQ4_K_M quantized, ~15 tok/s, 32k context
- SWEET SPOT
Dual RTX 6000 Ada Generation (2x 48GB) Workstation
96 GB VRAM
~$14000 NEWNVFP4 / Q4, ~28 tok/s, 128k context
- FULL POWER
Quad RTX 6000 Ada Generation (4x 48GB) Workstation
192 GB VRAM
~$28000 NEWFP8 / Q8, ~45 tok/s, 256k context
LEARN HOW TO SET IT UP
WILL MY BOX RUN LAGUNA S 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 S 2.1 takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMApple Mac Studio M2 Ultra (128GB Unified Memory) | SWEET SPOTDual RTX 6000 Ada Generation (2x 48GB) Workstation | FULL POWERQuad RTX 6000 Ada Generation (4x 48GB) Workstation |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~33 S | ~18 S | ~11 S |
| Summarize a documenta long report boiled down to the points that matter | ~1.7 MIN | ~54 S | ~33 S |
| Build a websitea small landing page, markup and styles together | ~8.9 MIN★★★★★ | ~4.8 MIN★★★★★ | ~3 MIN★★★★★ |
| Build a backendan API with routes, storage and tests | ~28 MIN★★★★★ | ~15 MIN★★★★★ | ~9.3 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~17 MIN★★★★★ | ~8.9 MIN★★★★★ | ~5.6 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
The internet is split on Laguna S 2.1. Biggest praise: speed. Biggest gripe: saying no too much.
80 POSTS · 14 COMMENTS · DEV.TO · X · DEV FORUMS · REDDIT · GITHUB · BLUESKY · BLOG
PEOPLE LOVE
- Answers come back fast (4 mentions)
PEOPLE COMPLAIN ABOUT
- Coding help disappoints some people (3 mentions)
- Refuses or lectures more than people like (3 mentions)
TOPIC BY TOPIC
- Help with codeMIXED FEELINGS
3 thumbs up · 3 thumbs down
- SpeedPEOPLE LOVE THIS
4 thumbs up · 0 thumbs down
- Saying no too muchA SORE SPOT
0 thumbs up · 3 thumbs down
LOOKING FOR AN ALTERNATIVE?
Other models the crowd has fully reviewed, starting with text models like this one.
GLM 5.3
DeepSeek V4 Flash> SOURCES (12)
- REDDIT · 2026-08-04 · 267 UPVOTESA llama.cpp PR caches “hot” MoE experts on the GPU — 33 → 56 tok/s reported with 8GB VRAM
A llama.cpp PR caches “hot” MoE experts on the GPU — 33 → 56 tok/s reported with 8GB VRAM
- REDDIT · 2026-07-22 · 227 UPVOTESUnsloth Quantization of Laguna S 2.1 Is Out
Unsloth Quantization of Laguna S 2.1 Is Out
- REDDIT · 2026-07-24 · 178 UPVOTESLaguna s.2.1 updated 2 hours ago. A post to show appreciation for the work they are doing.
Laguna s.2.1 updated 2 hours ago. A post to show appreciation for the work they are doing.
- REDDIT · 2026-07-22 · 144 UPVOTESDespite not being trained to, it turns out the Pearson correlation between a models AA Intelligence Index score and its ability to generate…
Despite not being trained to, it turns out the Pearson correlation between a models AA Intelligence Index score and its ability to generate Base64 encoded responses is 0.91
- REDDIT · 2026-07-22 · 141 UPVOTESToday was the perfect day for Poolside to drop Laguna S 2.1 because I just got these in! Finally have a half decent amount of VRAM. 3x V620…
Today was the perfect day for Poolside to drop Laguna S 2.1 because I just got these in! Finally have a half decent amount of VRAM. 3x V620 = 96 GB.
- REDDIT · 2026-07-22 · 124 UPVOTESLaguna S 2.1 looping fix incoming
Laguna S 2.1 looping fix incoming
- REDDIT · 2026-08-07 · 103 UPVOTESllama.cpp PR reports up to 169% faster quantized-KV decode at 118K context on Intel Battlemage from one SYCL kernel switch
llama.cpp PR reports up to 169% faster quantized-KV decode at 118K context on Intel Battlemage from one SYCL kernel switch
- REDDIT · 2026-07-24 · 86 UPVOTESAsking Laguna S 2.1: "I want to wash my car. The car wash is 69 meters away. Should I walk or drive?"
Asking Laguna S 2.1: "I want to wash my car. The car wash is 69 meters away. Should I walk or drive?"
- REDDIT · 2026-07-24 · 66 UPVOTESI'm impressed by Laguna S 2.1
I'm impressed by Laguna S 2.1
- REDDIT · 2026-07-31 · 51 UPVOTESWe've gotten some great medium sized models lately (DSV4 Flash 0731, Inkling Small, Laguna S 2.1, Step 3.7 Flash) but does anybody else…
We've gotten some great medium sized models lately (DSV4 Flash 0731, Inkling Small, Laguna S 2.1, Step 3.7 Flash) but does anybody else want to see some new 70-80b contenders?
- REDDIT · 2026-07-22 · 40 UPVOTESHow are we feeling about Poolside's Laguna S 2.1? (only comment if you've used it)
How are we feeling about Poolside's Laguna S 2.1? (only comment if you've used it)
- REDDIT · 2026-08-11 · 38 UPVOTESI tested the CMP170HX
I tested the CMP170HX