PrimeIntellectLlamaForCausalLM

INTELLECT 1 Instruct — Hardware Requirements & GPU Compatibility

Chat

INTELLECT-1-Instruct is the instruction-tuned version of INTELLECT-1, a 10-billion-parameter language model that Prime Intellect describes as the first collaboratively, globally distributed pretraining run of its scale: 1 trillion tokens of English text and code trained across up to 14 concurrent nodes on three continents contributed by 30 independent community participants, using the DiLoCo distributed-training algorithm. Post-training was handled by Arcee AI through supervised fine-tuning, eight rounds of direct preference optimization, and model merging, using the Llama-3 tokenizer and logits distilled from Llama-3.1-405B to improve alignment. At 10B parameters it fits a single consumer GPU once quantized, more comfortably at full precision on a higher-end card. Context length is 8,192 tokens. It is released under the Apache 2.0 license, permitting unrestricted commercial and research use, and was published in November 2024, alongside the base INTELLECT-1 model.

10.2B params · dense8K contextApache 2.0Nov 2024717 downloads125 likes5.3K quant downloads

Based on INTELLECT 1

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 8 GB (Q4_K_M · 8K)Best on 24 GB (Q4_K_M · ~97 tok/s · 8K)Macs from 16 GB (Q4_K_M · ~21 tok/s)

Apache 2.0

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run INTELLECT 1 Instruct

One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.

Ways to runOther ways

each opens step-by-step

Run in cloud

Runs on RTX 3080 (10 GB) (2 GB headroom) · Q4_K_M

Generation speed: ~73 tok/s; generation speed
Cost per 1M output tokens: $0.41; per 1M output tokens

Deployon Vast.aiBest valuefrom $0.11/hr

Compare GPUs →

Run INTELLECT 1 Instruct

7.8 GB VRAM at 8K context · fits the RTX 4090 (16.2 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/bartowski/INTELLECT-1-Instruct-GGUF:Q4_K_M
Raise the context window
Type it inside the chat. For the API server instead, start it with OLLAMA_CONTEXT_LENGTH=8192 ollama serve.
$/set parameter num_ctx 8192
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/bartowski/INTELLECT-1-Instruct-GGUF, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get https://huggingface.co/bartowski/INTELLECT-1-Instruct-GGUF@Q4_K_M
Open on Hugging Face ↗
Load and chat
Open the model loader, select the model and set Context Length to 8192. 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 bartowski/INTELLECT-1-Instruct-GGUF:Q4_K_M -c 8192 -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/bartowski/INTELLECT-1-Instruct-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 bartowski/INTELLECT-1-Instruct-GGUF:Q4_K_M -c 8192 -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 PrimeIntellect/INTELLECT-1-Instruct --max-model-len 8192
Runs the original weights, not the Q4_K_M GGUF; needs ~22 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/INTELLECT-1-Instruct-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/INTELLECT-1-Instruct-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.
$bartowski/INTELLECT-1-Instruct-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 3080 (10 GB)CheapestCheapest2 GB free	~73 tok/s	$0.41	Deploy on Vast.aifrom $0.11/hr (opens in new tab)	10 GB · 2 GB free · ~73 tok/s · $0.41 / 1M tokens
RTX 3060 12GB4 GB free	~35 tok/s	$0.50	Deploy on Vast.aifrom $0.061/hr (opens in new tab)	4 GB free · ~35 tok/s · $0.50 / 1M tokens
RTX A4000 (16 GB)8 GB free	~43 tok/s	$0.60	Deploy on Vast.aifrom $0.092/hr (opens in new tab)	16 GB · 8 GB free · ~43 tok/s · $0.60 / 1M tokens
RTX 4070 SUPER (12 GB)4 GB free	~48 tok/s	$0.59	Deploy on Vast.aifrom $0.10/hr (opens in new tab)	12 GB · 4 GB free · ~48 tok/s · $0.59 / 1M tokens
RTX 4000 Ada (20 GB)12 GB free	~35 tok/s	$1.61	Deploy on RunPodfrom $0.20/hr (opens in new tab)	20 GB · 12 GB free · ~35 tok/s · $1.61 / 1M tokens
RTX PRO 4500 (32 GB)24 GB free	~86 tok/s	$1.10	Deploy on RunPodfrom $0.34/hr (opens in new tab)	32 GB · 24 GB free · ~86 tok/s · $1.10 / 1M tokens
RTX 4090 (24 GB)16 GB free	~97 tok/s	$0.98	Deploy on RunPodfrom $0.34/hr (opens in new tab)	24 GB · 16 GB free · ~97 tok/s · $0.98 / 1M tokens
A40 (48 GB)40 GB free	~67 tok/s	$1.46	Deploy on RunPodfrom $0.35/hr (opens in new tab)	48 GB · 40 GB free · ~67 tok/s · $1.46 / 1M tokens

Compare GPUs →

How Much VRAM Does INTELLECT 1 Instruct Need?

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

VRAM	+32K	+128K	File	Quality	On RTX 4090 (reference)
3.5 GB	—	—	3 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Fits: 8K context · ~189 tok/s
3.7 GB	—	—	3.3 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Fits: 8K context · ~176 tok/s
3.8 GB	—	—	3.4 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Fits: 8K context · ~171 tok/s
4.1 GB	—	—	3.7 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Fits: 8K context · ~160 tok/s
4.6 GB	—	—	4.1 GB	3-bit · i-quant — Importance-weighted 3-bit	Fits: 8K context · ~142 tok/s
4.7 GB	—	—	3.7 GB	2-bit · small K-quant — 2-bit small K-quant	Fits: 8K context · ~138 tok/s
4.9 GB	—	—	4.4 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Fits: 8K context · ~135 tok/s
5.0 GB	—	—	4.6 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Fits: 8K context · ~131 tok/s
5.0 GB	—	—	4 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Fits: 8K context · ~131 tok/s
5.1 GB	—	—	4.6 GB	3-bit small — 3-bit small quantization	Fits: 8K context · ~128 tok/s
5.3 GB	—	—	4.8 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Fits: 8K context · ~125 tok/s
5.6 GB	—	—	5.1 GB	3-bit medium — 3-bit medium quantization	Fits: 8K context · ~116 tok/s
5.8 GB	—	—	5.9 GB	4-bit · legacy — 4-bit legacy quantization	Fits: 8K context · ~114 tok/s
5.9 GB	—	—	5.5 GB	3-bit large — 3-bit large quantization	Fits: 8K context · ~111 tok/s
6.1 GB	—	—	5.6 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Fits: 8K context · ~107 tok/s
6.4 GB	—	—	6.5 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Fits: 8K context · ~102 tok/s
6.4 GB	—	—	5.9 GB	4-bit small — 4-bit small quantization	Fits: 8K context · ~102 tok/s
6.4 GB	—	—	5.9 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Fits: 8K context · ~102 tok/s
6.8 GB	—	—	6.2 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Fits: 8K context · ~97 tok/s
6.9 GB	—	—	6.6 GB	4-bit large — 4-bit large quantization	Fits: 8K context · ~95 tok/s
7.7 GB	—	—	7.1 GB	5-bit small — 5-bit small quantization	Fits: 8K context · ~85 tok/s
7.9 GB	—	—	7.3 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Fits: 8K context · ~83 tok/s
8.1 GB	—	—	7.6 GB	5-bit large — 5-bit large quantization	Fits: 8K context · ~81 tok/s
9.1 GB	—	—	8.4 GB	6-bit, very good — 6-bit quantization, very good quality	Fits: 8K context · ~72 tok/s
10.9 GB	—	—	10.9 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Fits: 8K context · ~60 tok/s
21.1 GB	—	—	~20.4 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Fits: 8K context · ~31 tok/s

est.calculated, no published file yet

Selected Q4_K_M

4-bit medium quantization

7.8GBat 8K context

Weights: 6.1 GB
KV cache: 1.4 GB
Overhead: 0.3 GB

Fits RTX 4090 (reference) · 16.2 GB free · ~97 tok/s

≥50 tok/s needs ≥521 GB/s · RTX 5090 ~172, RTX 3090 Ti ~97

Which Hardware Can Run INTELLECT 1 Instruct?

Q4_K_M · 6.8 GB

INTELLECT 1 Instruct (Q4_K_M) requires 6.8 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 9+ GB is recommended. Using the full 8K context window can add up to 1.1 GB, bringing total usage to 7.8 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3080. 151 devices with unified memory can run INTELLECT 1 Instruct, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), NVIDIA Jetson Orin Nano 8GB (Super).

By VRAM tier · typical GPUs and Macs at Q4_K_M

Typical hardwareQuantVRAMtok/sContext that fits

8 GB · RTX 4060Q4_K_M · 7.8 GB · ~26 tok/s · 8K7.8~268K8 GB tier →

12 GB · RTX 3060 12GBQ4_K_M · 7.8 GB · ~35 tok/s · 8K7.8~358K12 GB tier →

16 GB · RTX 5060 Ti 16GBQ4_K_M · 7.8 GB · ~43 tok/s · 8K7.8~438K16 GB tier →

24 GB · RTX 4090referenceQ4_K_M · 7.8 GB · ~97 tok/s · 8K7.8~978K24 GB tier →

Mac 32 GB · Mac Mini M4Q4_K_M · 7.8 GB · ~12 tok/s · 8K7.8~128KMacs →

Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 7.8 GB · ~56 tok/s · 8K7.8~568KMacs →

How good is INTELLECT 1 Instruct?

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

MMLU
Epoch AI · Nov 2024Lab-reported
49.9%
#63/85 open
Lab-reported
BIG-Bench Hard
Epoch AI · Nov 2024Lab-reported
34.8%
#33/38 open
Lab-reported
GSM8K
Epoch AI · Nov 2024Lab-reported
38.6%
#36/62 open
Lab-reported
HellaSwag
Epoch AI · Nov 2024Lab-reported
71.4%
#29/47 open
Lab-reported
ARC Challenge
Epoch AI · Nov 2024Lab-reported
54.5%
#28/56 open
Lab-reported
WinoGrande
Epoch AI · Nov 2024Lab-reported
65.8%
#32/51 open
Lab-reported

All 6 INTELLECT 1 Instruct benchmarks

Alternatives to INTELLECT 1 Instruct

Similar models

6tok/s on RTX 4090

Model	Score	VRAM	tok/s	Why it is similar
Qwen3.5 9BAlibaba	99.4	6.6 GB	~103	same size
Ornith 1.5 9Bornith-ai	—	6.6 GB	~103	same size
GLM 4.1V 9B ThinkingZ.ai	—	7.2 GB	~100	same size
Qwen2.5 Omni 7BAlibaba	—	9.9 GB	~93	same size
Llama 3.2 11B Vision InstructMeta	—	9.8 GB	~93	same size
SOLAR 10.7B Instruct v1.0Upstage	—	7.5 GB	~92	same size

Specifications

Publisher: PrimeIntellect
Parameters: 10.2B
Architecture: LlamaForCausalLM
Context Length: 8,192 tokens
Vocabulary Size: 128,256
Release Date: 2024-11-28
License: Apache 2.0

Resources

Hugging Face model card (opens in a new tab)

Frequently Asked Questions

How much VRAM does INTELLECT 1 Instruct need?

INTELLECT 1 Instruct requires 6.8 GB of VRAM at Q4_K_M, or 21.1 GB at BF16. Full 8K context adds up to 1.1 GB (7.8 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 10.2B × 4.8 bits ÷ 8 = 6.1 GB

KV Cache + Overhead ≈ 0.7 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 ≈ 1.7 GB (at full 8K context)

VRAM usage by quantization

Q4_K_M

6.8 GB

Q4_K_M + full context

7.8 GB

Learn more about VRAM estimation →

What's the best quantization for INTELLECT 1 Instruct?

For INTELLECT 1 Instruct, Q4_K_M (6.8 GB) offers the best balance of quality and VRAM usage. Q4_K_L (6.9 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 3.5 GB.

VRAM requirement by quantization

IQ2_XXS

3.5 GB~53%

IQ3_XS

4.9 GB~73%

Q3_K_L

5.9 GB~86%

Q4_K_M ★

6.8 GB~89%

Q4_K_L

6.9 GB~90%

BF16

21.1 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run INTELLECT 1 Instruct on a Mac?

Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run INTELLECT 1 Instruct. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 6.8 GB of usable unified memory (RAM minus macOS overhead).

Can I run INTELLECT 1 Instruct locally?

Yes — INTELLECT 1 Instruct can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is INTELLECT 1 Instruct?

At Q4_K_M, INTELLECT 1 Instruct can reach ~708 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~97 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = (bandwidth GB/s ÷ model GB) × efficiency

Example: NVIDIA B200 → 8000 ÷ 6.78 × 0.65 = ~767 tok/s

Estimated speed at Q4_K_M (6.8 GB)

NVIDIA B200

~767 tok/s

NVIDIA GeForce RTX 4090

~97 tok/s

NVIDIA B300

~767 tok/s

AMD Instinct MI350X

~708 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 INTELLECT 1 Instruct?

At Q4_K_M, the download is about 6.13 GB. The full-precision BF16 version is 20.42 GB. The smallest option (IQ2_XXS) is 2.81 GB.

Which GPUs can run INTELLECT 1 Instruct?

52 consumer GPUs can run INTELLECT 1 Instruct at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT, AMD Radeon RX 7600. 38 GPUs have plenty of headroom for comfortable inference.

Which devices can run INTELLECT 1 Instruct?

151 devices with unified memory can run INTELLECT 1 Instruct at Q4_K_M (6.8 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.