poolsideLagunaForCausalLM

Laguna S 2.1 — Hardware Requirements & GPU Compatibility

ChatFunctions

Laguna S 2.1 is a 117.6B-parameter open language model from poolside. It supports a context window of up to 1,048,576 tokens. At Q4_K_M it needs about 71.04 GB of VRAM — see which GPUs and Macs can run it below.

117.6B params · MoE8.4B active1049K contextopenmdw-1.1Jul 2026343.6K downloads1.1K likes631.1K 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.

Laguna S 2.1 runs on a 96 GB+ Mac or a workstation GPU

Best local pick → Mac Studio (M3 Ultra, 96GB) · Q4_K_M · 72.5 GB · ~54 tok/s

Macs from 48 GB (IQ2_M · ~61 tok/s)

openmdw-1.1

Not set up for it?

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run Laguna S 2.1

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 6000 (96 GB) (23 GB headroom) · Q4_K_M

Generation speed: ~142 tok/s; generation speed
Cost per 1M output tokens: $2.70; per 1M output tokens

Deployon Vast.aiBest optionfrom $1.38/hr

Compare GPUs →

Run Laguna S 2.1

72.5 GB VRAM at 16K context · needs offload on the RTX 4090 (48.5 GB short)

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-S-2.1-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 search poolside/laguna-s-2.1, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get poolside/laguna-s-2.1@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-S-2.1-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-S-2.1-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-S-2.1-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-S-2.1 --max-model-len 16384
Runs the original weights, not the Q4_K_M GGUF; needs ~259 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 poolside/Laguna-S-2.1-NVFP4-mlx
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 poolside/Laguna-S-2.1-NVFP4-mlx

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-S-2.1-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 6000 Max-Q (96 GB)CheapestCheapest23 GB free	~142 tok/s	$2.70	Deploy on Vast.aifrom $1.38/hr (opens in new tab)	96 GB · 23 GB free · ~142 tok/s · $2.70 / 1M tokens
RTX PRO 6000 (96 GB)23 GB free	~142 tok/s	$3.30	Deploy on RunPodfrom $1.69/hr (opens in new tab)	96 GB · 23 GB free · ~142 tok/s · $3.30 / 1M tokens
RTX PRO 6000 Server (96 GB)23 GB free	~132 tok/s	$3.55	Deploy on RunPodfrom $1.69/hr (opens in new tab)	96 GB · 23 GB free · ~132 tok/s · $3.55 / 1M tokens
H200 SXM (141 GB)68 GB free	~234 tok/s	$4.27	Deploy on RunPodfrom $3.59/hr (opens in new tab)	141 GB · 68 GB free · ~234 tok/s · $4.27 / 1M tokens
H200 NVL (141 GB)68 GB free	~234 tok/s	$4.80	Deploy on Vast.aifrom $4.04/hr (opens in new tab)	141 GB · 68 GB free · ~234 tok/s · $4.80 / 1M tokens
B200 (192 GB)119 GB free	~276 tok/s	$6.02	Deploy on RunPodfrom $5.98/hr (opens in new tab)	192 GB · 119 GB free · ~276 tok/s · $6.02 / 1M tokens

Compare GPUs →

How Much VRAM Does Laguna S 2.1 Need?

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

VRAM	+32K	+128K	File	Quality	On RTX 4090 (reference)
32.8 GB	35.9 GB	45.5 GB	37.2 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Does not fit: needs 34.2 GB
35.8 GB	38.8 GB	48.5 GB	35.6 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Does not fit: needs 37.2 GB
37.2 GB	40.3 GB	49.9 GB	36.3 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Does not fit: needs 38.7 GB
40.2 GB	43.2 GB	52.9 GB	37.3 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Does not fit: needs 41.6 GB
46.1 GB	49.1 GB	58.7 GB	44.3 GB	3-bit · i-quant — Importance-weighted 3-bit	Does not fit: needs 47.5 GB
49 GB	52 GB	61.7 GB	54 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Does not fit: needs 50.4 GB
50.5 GB	53.5 GB	63.2 GB	48.4 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Does not fit: needs 51.9 GB
50.5 GB	53.5 GB	63.2 GB	41.8 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Does not fit: needs 51.9 GB
51.9 GB	55 GB	64.6 GB	51.6 GB	3-bit small — 3-bit small quantization	Does not fit: needs 53.3 GB
53.4 GB	56.4 GB	66.1 GB	56.5 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Does not fit: needs 54.8 GB
57.8 GB	60.8 GB	70.5 GB	54 GB	3-bit medium — 3-bit medium quantization	Does not fit: needs 59.2 GB
59.3 GB	62.3 GB	72 GB	67 GB	4-bit · legacy — 4-bit legacy quantization	Does not fit: needs 60.7 GB
60.8 GB	63.8 GB	73.4 GB	56.2 GB	3-bit large — 3-bit large quantization	Does not fit: needs 62.2 GB
63.7 GB	66.7 GB	76.4 GB	57.6 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Does not fit: needs 65.1 GB
66.6 GB	69.7 GB	79.3 GB	58.7 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Does not fit: needs 68 GB
66.6 GB	69.7 GB	79.3 GB	68.6 GB	4-bit small — 4-bit small quantization	Does not fit: needs 68 GB
66.6 GB	69.7 GB	79.3 GB	74 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Does not fit: needs 68 GB
71.0 GB	74.1 GB	83.7 GB	96 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Does not fit: needs 72.5 GB
81.3 GB	84.4 GB	94 GB	82.7 GB	5-bit small — 5-bit small quantization	Does not fit: needs 82.7 GB
84.3 GB	87.3 GB	96.9 GB	87.9 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Does not fit: needs 85.7 GB
85.7 GB	88.8 GB	98.4 GB	84.2 GB	5-bit large — 5-bit large quantization	Does not fit: needs 87.1 GB
97.5 GB	100.5 GB	110.2 GB	97.9 GB	6-bit, very good — 6-bit quantization, very good quality	Does not fit: needs 98.9 GB
118.1 GB	121.1 GB	130.7 GB	128.8 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Does not fit: needs 119.5 GB
235.6 GB	238.7 GB	248.3 GB	~235.1 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Does not fit: needs 237 GB

Selected Q4_K_M

4-bit medium quantization

72.5GBat 16K context

Weights: 70.5 GB
KV cache: 1.6 GB
Overhead: 0.3 GB

Doesn't fit RTX 4090 (reference) · needs 48.4 GB more

≥50 tok/s needs ≥452 GB/s · M3 Ultra ~54, M2 Ultra ~53

Or rent RTX PRO 6000 from $1.38/hr.

Which Hardware Can Run Laguna S 2.1?

Q4_K_M · 71.0 GB

Laguna S 2.1 (Q4_K_M) requires 71.0 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 93+ GB is recommended. Using the full 1049K context window can add up to 102.9 GB, bringing total usage to 173.9 GB. No consumer GPU has enough memory. 30 devices with unified memory can run Laguna S 2.1, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 96GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M

Typical hardwareQuantVRAMtok/sContext that fits

8–16 GB · RTX 4060, RTX 3060 12GB, RTX 5060 Ti 16GBDoesn't fit — smallest quant needs 34.2 GB8 GB tier →12 GB tier →16 GB tier →

24 GB · RTX 4090referenceDoesn't fit — smallest quant needs 34.2 GB24 GB tier →

Mac 32 GB · Mac Mini M4Doesn't fit — smallest quant needs 34.2 GBMacs →

Mac 64 GB · Mac Studio M4 MaxQ3_K_M · 59.2 GB · ~49 tok/s · 16K59.2~4916KMacs →

Alternatives to Laguna S 2.1

Similar models

6tok/s on RTX 4090

Model	Score	VRAM	tok/s	Why it is similar
GPT OSS 120BOpenAI	97.5	70.8 GB	—needs > 24 GB	same size
GLM 4.5 AirZ.ai	96.3	67.6 GB	—needs > 24 GB	same size
C4ai Command A 03 2025Cohere	79.8	78.7 GB	—needs > 24 GB	same size
NVIDIA Nemotron 3.5 Super VL 120B A12B BF16NVIDIA	—	87.7 GB	—needs > 24 GB	same size
Ling 3.0 Flash VLInclusion AI	—	76 GB	—needs > 24 GB	same size
Leanstral 1.5 119B A6BMistral AI	—	84 GB	—needs > 24 GB	same size

Specifications

Publisher: poolside
Parameters: 117.6B
Active parameters: 8.4B per token
Architecture: LagunaForCausalLM
Context Length: 1,048,576 tokens
Vocabulary Size: 100,352
Release Date: 2026-07-13
License: openmdw-1.1

Resources

Hugging Face model card (opens in a new tab)

Frequently Asked Questions

How much VRAM does Laguna S 2.1 need?

Laguna S 2.1 requires 71.0 GB of VRAM at Q4_K_M, or 235.6 GB at BF16. Full 1049K context adds up to 102.9 GB (173.9 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 117.6B × 4.8 bits ÷ 8 = 70.5 GB

KV Cache + Overhead ≈ 0.5 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 ≈ 103.4 GB (at full 1049K context)

VRAM usage by quantization

Q4_K_M

71.0 GB

Q4_K_M + full context

173.9 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 5090 run Laguna S 2.1?

No — Laguna S 2.1 requires at least 32.8 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.

What's the best quantization for Laguna S 2.1?

For Laguna S 2.1, Q4_K_M (71.0 GB) offers the best balance of quality and VRAM usage. Q5_K_S (81.3 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 32.8 GB.

VRAM requirement by quantization

IQ2_XXS

32.8 GB~53%

IQ3_S

50.5 GB~75%

Q3_K_L

60.8 GB~86%

Q4_K_M ★

71.0 GB~89%

Q5_K_S

81.3 GB~92%

BF16

235.6 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Laguna S 2.1 on a Mac?

Yes — Mac Studio M5 Max (48 GB) and 55 other Macs can run Laguna S 2.1, the smallest at IQ2_M. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 71.0 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 96GB).

Can I run Laguna S 2.1 locally?

Yes — Laguna S 2.1 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Laguna S 2.1?

At Q4_K_M, Laguna S 2.1 can reach ~94 tok/s on AMD Instinct MI350X. 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 → 5.1 GB active ÷ (8000 × 0.65) = 0.98 ms, plus 48 layers × 0.055 ms = 2.64 ms, so 1000 ÷ 3.62 ms = ~276 tok/s

Estimated speed at Q4_K_M (71.0 GB)

NVIDIA B200

~276 tok/s

NVIDIA B300

~276 tok/s

NVIDIA GH200 Grace Hopper Superchip

~236 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 S 2.1?

At Q4_K_M, the download is about 70.54 GB. The full-precision BF16 version is 235.12 GB. The smallest option (IQ2_XXS) is 32.33 GB.

Which GPUs can run Laguna S 2.1?

No single consumer GPU has enough VRAM to run Laguna S 2.1 at Q4_K_M (71.0 GB). Multi-GPU or professional hardware is required.

Which devices can run Laguna S 2.1?

30 devices with unified memory can run Laguna S 2.1 at Q4_K_M (71.0 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (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.