poolsideLagunaForCausalLM

Laguna XS.2 — Hardware Requirements & GPU Compatibility

ChatFunctions

Laguna XS.2 is a 33.4B-parameter open language model from poolside. It supports a context window of up to 262,144 tokens. At Q4_K_M it needs about 20.48 GB of VRAM — see which GPUs and Macs can run it below.

33.4B params · MoE3B active262K contextApache 2.0Apr 2026493.5K downloads320 likes2.9K quant downloads

Your hardware

Verdicts below are for the reference RTX 4090. Set yours to personalise the page.

Detection reads your browser's GPU name with no data sent anywhere.

Runs great on a 24 GB GPU

Runs from 16 GB (Q2_K · 16K)Best on 24 GB (Q4_K_M · ~199 tok/s · 64K)Macs from 24 GB (Q4_K_M · ~60 tok/s)

Apache 2.0

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run Laguna XS.2

One recommended way. Every other method, OS and quant is in the step-by-step guide.

Other ways

each opens step-by-step

Run in cloud

Fits on RTX PRO 4500 (32 GB) (10 GB headroom) · Q4_K_M

Generation speed: ~186 tok/s; generation speed
Cost per 1M output tokens: $0.51; per 1M output tokens

Deployon RunPodBest optionfrom $0.34/hr

Compare GPUs →

Run Laguna XS.2

21.3 GB VRAM at 16K context · fits the RTX 4090 (2.7 GB free)

Works on macOS, Windows and Linux.

Install Ollama (Linux)
$curl -fsSL https://ollama.com/install.sh | sh
Install Ollama (macOS / Windows)
Download and run the app; it installs the ollama command.
Download Ollama ↗
Run Q4_K_M
$ollama run hf.co/poolside-laguna-hackathon/Laguna-XS.2-GGUF:Q4_K_M
Raise the context window
Type it inside the chat. For the API server instead, start it with OLLAMA_CONTEXT_LENGTH=16384 ollama serve.
$/set parameter num_ctx 16384
A larger context needs more memory than the weights alone.
Use the API
Ollama serves its API on http://localhost:11434.

Works on macOS, Windows and Linux.

Install LM Studio
Desktop app for macOS (Apple silicon), Windows and Linux.
lmstudio.ai ↗
Download the model
In the app press Ctrl/Cmd + Shift + M and paste https://huggingface.co/poolside-laguna-hackathon/Laguna-XS.2-GGUF, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get https://huggingface.co/poolside-laguna-hackathon/Laguna-XS.2-GGUF@Q4_K_M
Open on Hugging Face ↗
Load and chat
Open the model loader, select the model and set Context Length to 16384. Leave GPU offload at maximum if it fits.
Start the local server (optional)
$lms server start

Works on macOS, Windows and Linux.

Install llama.cpp (macOS)
MacPorts, Nix and conda-forge also work.
$brew install llama.cpp
Install options ↗
Install llama.cpp (Windows)
$winget install llama.cpp
Install llama.cpp (Linux)
Download the prebuilt llama-*-bin-ubuntu-* archive for your GPU backend (CPU, Vulkan) from the releases page, unpack it and run the binaries from that folder. Homebrew (brew install llama.cpp) and conda-forge also work on Linux.
llama.cpp releases ↗
Start the server
Downloads the model on first run. The chat UI and an OpenAI-compatible API are at http://localhost:8080.
$llama-server -hf poolside-laguna-hackathon/Laguna-XS.2-GGUF:Q4_K_M -c 16384 -ngl 99 --port 8080

Works on macOS, Windows and Linux.

Option A: Docker Model Runner
Docker Desktop: Settings > AI > Enable Docker Model Runner. Docker Engine on Linux: install the docker-model-plugin package.
$docker model run hf.co/poolside-laguna-hackathon/Laguna-XS.2-GGUF:Q4_K_M
Model Runner docs ↗
The quant tag after the repo is how Hugging Face cards write it; if your Docker version rejects it, drop the tag and pick the quant by repo.
Option B: llama.cpp server image (NVIDIA)
Needs the NVIDIA Container Toolkit. Models are cached in the llama-cache volume. Open http://localhost:8080.
$docker run --gpus all -p 8080:8080 -v llama-cache:/root/.cache ghcr.io/ggml-org/llama.cpp:server-cuda -hf poolside-laguna-hackathon/Laguna-XS.2-GGUF:Q4_K_M -c 16384 -ngl 99 --host 0.0.0.0 --port 8080
Image tags (cuda13, vulkan, rocm) ↗
Cache path inside the image is not in the docs; remove the -v flag if the volume complains.

Linux only.

Install vLLM
Linux only; on Windows run it inside WSL2. NVIDIA GPU shown; AMD ROCm, Intel and CPU builds have their own install pages.
$uv venv --python 3.12 --seed
$source .venv/bin/activate
$uv pip install vllm --torch-backend=auto
vLLM installation ↗
Serve the model
OpenAI-compatible API at http://localhost:8000/v1.
$vllm serve poolside/Laguna-XS.2 --max-model-len 16384
Runs the original weights, not the Q4_K_M GGUF; needs ~74 GB of GPU memory.

macOS only.

MLX supports only macOS, so follow these steps for that setup.

Install mlx-lm
$pip install mlx-lm
Chat in the terminal
$mlx_lm.chat --model mlx-community/Laguna-XS.2-4bit
Downloads the model on first run. The MLX repo has its own quantisation; it may not match the quant selected above.
Or start a local server
Listens on http://localhost:8080.
$mlx_lm.server --model mlx-community/Laguna-XS.2-4bit

Works on macOS, Windows and Linux.

Install Jan
jan.ai ↗
Find the model
Open the Hub tab, paste this model ID in Search, and choose Q4_K_M if offered.
$poolside-laguna-hackathon/Laguna-XS.2-GGUF
Use the local API (optional)
Settings > Local API Server > Start Server. It listens on http://127.0.0.1:1337.

Rent GPUs hourly to run Q4_K_M right there. Pricing comes from partner marketplaces and varies.

Available GPU rentals for Q4_K_M, prioritized by value
GPU	Speed	$/1M tokens	Rent
RTX PRO 4500 (32 GB)CheapestCheapest10 GB free	~186 tok/s	$0.51	Deploy on RunPodfrom $0.34/hr (opens in new tab)	32 GB · 10 GB free · ~186 tok/s · $0.51 / 1M tokens
A40 (48 GB)26 GB free	~159 tok/s	$0.61	Deploy on RunPodfrom $0.35/hr (opens in new tab)	48 GB · 26 GB free · ~159 tok/s · $0.61 / 1M tokens
A40 (48 GB)26 GB free	~159 tok/s	$0.70	Deploy on Vast.aifrom $0.40/hr (opens in new tab)	48 GB · 26 GB free · ~159 tok/s · $0.70 / 1M tokens
RTX 5090 (32 GB)10 GB free	~264 tok/s	$0.73	Deploy on RunPodfrom $0.69/hr (opens in new tab)	32 GB · 10 GB free · ~264 tok/s · $0.73 / 1M tokens
RTX 6000 Ada (48 GB)26 GB free	~194 tok/s	$1.05	Deploy on Vast.aifrom $0.73/hr (opens in new tab)	48 GB · 26 GB free · ~194 tok/s · $1.05 / 1M tokens
RTX 6000 Ada (48 GB)26 GB free	~194 tok/s	$1.06	Deploy on RunPodfrom $0.74/hr (opens in new tab)	48 GB · 26 GB free · ~194 tok/s · $1.06 / 1M tokens
RTX 5090 (32 GB)10 GB free	~264 tok/s	$0.80	Deploy on Vast.aifrom $0.76/hr (opens in new tab)	32 GB · 10 GB free · ~264 tok/s · $0.80 / 1M tokens
L40S (48 GB)26 GB free	~182 tok/s	$1.22	Deploy on Vast.aifrom $0.80/hr (opens in new tab)	48 GB · 26 GB free · ~182 tok/s · $1.22 / 1M tokens

Compare GPUs →

How Much VRAM Does Laguna XS.2 Need?

Select a quantization level; the listed hardware adheres to it.

VRAM	+32K	+128K	File	Quality	On RTX 4090 (reference)
14.6 GB	16.3 GB	21.7 GB	~14.2 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Fits: 128K context · ~237 tok/s
16.7 GB	18.4 GB	23.8 GB	~16.3 GB	3-bit medium — 3-bit medium quantization	Fits: 128K context · ~222 tok/s
20.5 GB	22.2 GB	27.5 GB	~20.1 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Fits: 64K context · ~199 tok/s
24.2 GB	25.9 GB	31.3 GB	~23.8 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Does not fit: needs 25 GB
28 GB	29.7 GB	35.1 GB	~27.6 GB	6-bit, very good — 6-bit quantization, very good quality	Does not fit: needs 28.8 GB
33.9 GB	35.5 GB	40.9 GB	~33.4 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Does not fit: needs 34.6 GB
67.3 GB	69 GB	74.4 GB	66.9 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Does not fit: needs 68.1 GB

est.calculated, no published file yet

Selected Q4_K_M

4-bit medium quantization

21.3GBat 16K context

Weights: 20.1 GB
KV cache: 0.9 GB
Overhead: 0.3 GB

Fits RTX 4090 (reference) · 2.7 GB free · ~199 tok/s

≥50 tok/s needs ≥160 GB/s · RTX 5090 ~264, RTX 3090 Ti ~199

Which Hardware Can Run Laguna XS.2?

Q4_K_M · 20.5 GB

Laguna XS.2 (Q4_K_M) requires 20.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 27+ GB is recommended. Using the full 262K context window can add up to 14.2 GB, bringing total usage to 34.7 GB. 7 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 122 devices with unified memory can run Laguna XS.2, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M1 Max (32 GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M

Typical hardwareQuantVRAMtok/sContext that fits

8–12 GB · RTX 4060, RTX 3060 12GBDoesn't fit — smallest quant needs 15.4 GB8 GB tier →12 GB tier →

16 GB · RTX 5060 Ti 16GBQ2_K · 15.4 GB · ~149 tok/s · 16K15.4~14916K16 GB tier →

24 GB · RTX 4090referenceQ4_K_M · 21.3 GB · ~199 tok/s · 64K21.3~19964K24 GB tier →

Mac 32 GB · Mac Mini M4Q4_K_M · 21.3 GB · ~33 tok/s · 128K21.3~33128KMacs →

Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 21.3 GB · ~78 tok/s · 256K21.3~78256KMacs →

How good is Laguna XS.2?

llmrun Score blends public benchmark boards into one number; 100 = the average open model.

Not enough data yet for an llmrun Score.

A Score needs results on enough independent boards. How the Composite Score is built

Benchmarks

LMArena WebDev
LMArena · Oct 2026Arena
1302.5
#43/52 open
Arena
ProofBench
Epoch AI · Apr 2026Lab-reported
0.0%
#25/26 open
Lab-reported

All 2 Laguna XS.2 benchmarks

Alternatives to Laguna XS.2

Similar models

6tok/s on RTX 4090

Model	Score	VRAM	tok/s	Why it is similar
Qwen3.6 35B A3BAlibaba	106.8	22 GB	~151	same size
MiniCPM V 4.7 35B A3BOpenBMB	—	22.1 GB	~169	same size
Ornith 1.5 35B A3Bornith-ai	—	22 GB	~151	same size
NVIDIA Nemotron 3.5 Lightning 30B A3B BF16NVIDIA	—	19.3 GB	~197	same size
Agents A1InternScience	—	21.5 GB	~171	same size
Intern S2 PreviewInternLM	—	73.2 GB	—needs > 24 GB	same size

Specifications

Publisher: poolside
Parameters: 33.4B
Active parameters: 3B per token
Architecture: LagunaForCausalLM
Context Length: 262,144 tokens
Vocabulary Size: 100,352
Release Date: 2026-04-23
License: Apache 2.0

Resources

Hugging Face model card (opens in a new tab)

Frequently Asked Questions

How much VRAM does Laguna XS.2 need?

Laguna XS.2 requires 20.5 GB of VRAM at Q4_K_M, or 67.3 GB at BF16. Full 262K context adds up to 14.2 GB (34.7 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 33.4B × 4.8 bits ÷ 8 = 20.1 GB

KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

KV Cache + Overhead ≈ 14.6 GB (at full 262K context)

VRAM usage by quantization

Q4_K_M

20.5 GB

Q4_K_M + full context

34.7 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 4090 run Laguna XS.2?

Yes, at Q4_K_M (20.5 GB) or lower. Higher quantizations like Q5_K_M (24.2 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.

What's the best quantization for Laguna XS.2?

For Laguna XS.2, Q4_K_M (20.5 GB) offers the best balance of quality and VRAM usage. Q5_K_M (24.2 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 14.6 GB.

VRAM requirement by quantization

Q2_K

14.6 GB~75%

Q4_K_M ★

20.5 GB~89%

Q5_K_M

24.2 GB~92%

Q6_K

28.0 GB~95%

Q8_0

33.9 GB~99%

BF16

67.3 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Laguna XS.2 on a Mac?

Yes — Mac Mini M5 Pro (24 GB) and 106 other Macs can run Laguna XS.2. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 20.5 GB of usable unified memory (RAM minus macOS overhead).

Can I run Laguna XS.2 locally?

Yes — Laguna XS.2 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Laguna XS.2?

At Q4_K_M, Laguna XS.2 can reach ~119 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~199 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)

Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.

Example: NVIDIA B200 → 1.8 GB active ÷ (8000 × 0.65) = 0.36 ms, plus 40 layers × 0.055 ms = 2.20 ms, so 1000 ÷ 2.56 ms = ~391 tok/s

Estimated speed at Q4_K_M (20.5 GB)

NVIDIA B200

~391 tok/s

NVIDIA GeForce RTX 4090

~199 tok/s

NVIDIA B300

~391 tok/s

NVIDIA GH200 Grace Hopper Superchip

~360 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of Laguna XS.2?

At Q4_K_M, the download is about 20.07 GB. The full-precision BF16 version is 66.89 GB. The smallest option (Q2_K) is 14.21 GB.

Which GPUs can run Laguna XS.2?

7 consumer GPUs can run Laguna XS.2 at Q4_K_M. Top options include NVIDIA GeForce RTX 5090, AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090. 1 GPU has plenty of headroom for comfortable inference.

Which devices can run Laguna XS.2?

122 devices with unified memory can run Laguna XS.2 at Q4_K_M (20.5 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.