TII UAEFalconFalconForCausalLM

Falcon 40B Instruct — Hardware Requirements & GPU Compatibility

Chat

Falcon-40B-Instruct is TII's 40-billion-parameter causal decoder-only chat model, fine-tuned from the Falcon-40B base model on the Baize chat dataset so it is ready to use as an assistant out of the box. Falcon-40B was, at release, the top-ranked fully open model on the Open LLM Leaderboard, and the architecture is optimized for inference with FlashAttention and multi-query attention. It is trained and used primarily in English, with some French capability, and TII recommends further safety evaluation before production use. At 40B parameters it needs a multi-GPU workstation in half precision, while a 4-bit quantization brings it within reach of a single high-memory GPU. License is Apache 2.0, permitting unrestricted commercial and research use. It was published in May 2023, one day after the Falcon-40B base model it is fine-tuned from.

40B params · denseApache 2.0May 202311.0K downloads1.2K likes6.1K 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.

C

Usable on a 24 GB GPU

Runs on 24 GB (IQ3_M · ~33 tok/s · 16K)Macs from 24 GB (IQ2_M · ~15 tok/s)

Apache 2.0

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run Falcon 40B 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.

Recommended

Ollama

one command · needs Ollama installed

$ollama run hf.co/mradermacher/falcon-40b-instruct-i1-GGUF:Q4_K_M

Q4_K_M · needs ~30.4 GB · 16K context

Download: falcon-40b-instruct.i1-Q4_K_M.gguf25.99 GBmradermacher

install, OS, other quants

Ways to runOther ways

each opens step-by-step

Run in cloud

Runs on A100 40GB (9 GB headroom) · Q4_K_M

Generation speed: ~38 tok/s; generation speed
Cost per 1M output tokens: $4.39; per 1M output tokens

Deployon Vast.aiBest valuefrom $0.61/hr

Compare GPUs →

or

How Much VRAM Does Falcon 40B Instruct Need?

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

Quantization	VRAM	+32K	+128K	File	Quality	On RTX 4090 (reference)
	12.1 GB	20.7 GB	48.3 GB	11.9 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Fits: 32K context · ~54 tok/s
	13.2 GB	21.8 GB	49.4 GB	13.1 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Fits: 32K context · ~50 tok/s
	13.8 GB	22.4 GB	50 GB	13.9 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Fits: 32K context · ~48 tok/s
	14.8 GB	23.5 GB	51.1 GB	15 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Fits: 32K context · ~44 tok/s
	17.1 GB	25.7 GB	53.3 GB	16.9 GB	3-bit · i-quant — Importance-weighted 3-bit	Fits: 16K context · ~38 tok/s
	18.1 GB	26.8 GB	54.4 GB	18.3 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Fits: 16K context · ~36 tok/s
	18.7 GB	27.3 GB	54.9 GB	16.3 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Fits: 16K context · ~35 tok/s
	18.7 GB	27.3 GB	54.9 GB	18.8 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Fits: 16K context · ~35 tok/s
	19.3 GB	27.9 GB	55.5 GB	18.8 GB	3-bit small — 3-bit small quantization	Fits: 16K context · ~34 tok/s
	19.8 GB	28.4 GB	56 GB	19.7 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Fits: 16K context · ~33 tok/s
	21.4 GB	30.1 GB	57.7 GB	20.6 GB	3-bit medium — 3-bit medium quantization	Fits: 8K context · ~31 tok/s
	22 GB	30.6 GB	58.2 GB	23.9 GB	4-bit · legacy — 4-bit legacy quantization	Fits: 8K context · ~30 tok/s
	22.6 GB	31.2 GB	58.8 GB	22.1 GB	3-bit large — 3-bit large quantization	Fits: 4K context · ~29 tok/s
	23.6 GB	32.3 GB	59.9 GB	22.5 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Fits: 2K context · ~28 tok/s
	24.8 GB	33.4 GB	61 GB	24.3 GB	4-bit small — 4-bit small quantization	Does not fit: needs 28.8 GB
	26.4 GB	35 GB	62.6 GB	26 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Does not fit: needs 30.4 GB
	30.3 GB	38.9 GB	66.5 GB	29.5 GB	5-bit small — 5-bit small quantization	Does not fit: needs 34.3 GB
	31.4 GB	40 GB	67.6 GB	31.1 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Does not fit: needs 35.4 GB
	36.3 GB	44.9 GB	72.5 GB	35 GB	6-bit, very good — 6-bit quantization, very good quality	Does not fit: needs 40.3 GB
	44 GB	52.6 GB	80.2 GB	45 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Does not fit: needs 48 GB
	88 GB	96.6 GB	124.2 GB	~80 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Does not fit: needs 92 GB

est.calculated, no published file yet

Selected Q4_K_M

4-bit medium quantization

30.4GBat 16K context

Weights: 25.5 GB
KV cache: 4.6 GB
Overhead: 0.3 GB

Doesn't fit RTX 4090 (reference) · needs 6.4 GB more, IQ3_M fits

≥30 tok/s needs ≥1.22 TB/s · RTX 5090 ~44

Or .

Which Hardware Can Run Falcon 40B Instruct?

Q4_K_M · 26.4 GB

Falcon 40B Instruct (Q4_K_M) requires 26.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 35+ GB is recommended. 1 GPU can run it, including NVIDIA GeForce RTX 5090. 97 devices with unified memory can run Falcon 40B Instruct, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M5 Max (36 GB).

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 16.1 GB8 GB tier →12 GB tier →16 GB tier →

24 GB · RTX 4090referenceIQ3_M · 23.8 GB · ~33 tok/s · 16K23.8~3316K24 GB tier →

Mac 32 GB · Mac Mini M4Q4_K_M · 30.4 GB · ~3 tok/s · 8Ktight30.4~38KMacs →

Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 30.4 GB · ~15 tok/s · 32K30.4~1532KMacs →

How good is Falcon 40B 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

GSM8K
Epoch AI · May 2023Lab-reported
33.8%
#43/62 open
Lab-reported

Falcon 40B Instruct benchmark results

Alternatives to Falcon 40B Instruct

Falcon family · 13 models →

Similar models

3tok/s on RTX 4090

Model	Score	VRAM	tok/s	Why it is similar
FlintRecursiveSI	—	23.5 GB	~205	same size
Phi 3.5 MoE InstructMicrosoft	—	23.9 GB	~141	same size
K2 Horizon MoVA 36B A4BIFM	—	23.4 GB	~162	same size

Specifications

Publisher: TII UAE
Family: Falcon
Parameters: 40B
Architecture: FalconForCausalLM
Vocabulary Size: 65,024
Release Date: 2023-05-25
License: Apache 2.0

Resources

Frequently Asked Questions

How much VRAM does Falcon 40B Instruct need?

Falcon 40B Instruct requires 26.4 GB of VRAM at Q4_K_M, or 88 GB at BF16.

VRAM = Weights + KV Cache + Overhead

Weights = 40B × 4.8 bits ÷ 8 = 24 GB

KV Cache + Overhead ≈ 2.4 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.

VRAM usage by quantization

Q4_K_M

26.4 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 4090 run Falcon 40B Instruct?

Yes, at IQ4_XS (23.6 GB) or lower. Higher quantizations like Q4_K_S (24.8 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.

What's the best quantization for Falcon 40B Instruct?

For Falcon 40B Instruct, Q4_K_M (26.4 GB) offers the best balance of quality and VRAM usage. Q5_K_S (30.3 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 12.1 GB.

VRAM requirement by quantization

IQ2_XXS

12.1 GB~53%

IQ3_XS

18.1 GB~73%

Q3_K_M

21.4 GB~83%

Q4_K_M ★

26.4 GB~89%

Q5_K_S

30.3 GB~92%

BF16

88.0 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Falcon 40B Instruct on a Mac?

Yes — Mac Mini M5 Pro (24 GB) and 106 other Macs can run Falcon 40B Instruct, 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 26.4 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio M1 Max (32 GB).

Can I run Falcon 40B Instruct locally?

Yes — Falcon 40B Instruct can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Falcon 40B Instruct?

At Q4_K_M, Falcon 40B Instruct can reach ~182 tok/s on AMD Instinct MI350X. 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 ÷ 26.4 × 0.65 = ~197 tok/s

Estimated speed at Q4_K_M (26.4 GB)

NVIDIA B200

~197 tok/s

NVIDIA B300

~197 tok/s

AMD Instinct MI350X

~182 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 Falcon 40B Instruct?

At Q4_K_M, the download is about 24.00 GB. The full-precision BF16 version is 80.00 GB. The smallest option (IQ2_XXS) is 11.00 GB.

Which GPUs can run Falcon 40B Instruct?

1 consumer GPU can run Falcon 40B Instruct at Q4_K_M. Top options include NVIDIA GeForce RTX 5090.

Which devices can run Falcon 40B Instruct?

97 devices with unified memory can run Falcon 40B Instruct at Q4_K_M (26.4 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.

Run Falcon 40B Instruct

30.4 GB VRAM at 16K context

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/mradermacher/falcon-40b-instruct-i1-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/mradermacher/falcon-40b-instruct-i1-GGUF, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get https://huggingface.co/mradermacher/falcon-40b-instruct-i1-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 mradermacher/falcon-40b-instruct-i1-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/mradermacher/falcon-40b-instruct-i1-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 mradermacher/falcon-40b-instruct-i1-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 tiiuae/falcon-40b-instruct --max-model-len 16384
Runs the original weights, not the Q4_K_M GGUF; needs ~88 GB of GPU memory.

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.
$mradermacher/falcon-40b-instruct-i1-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
A100 40GBCheapestCheapest9 GB free	~38 tok/s	$4.39	Deploy on Vast.aifrom $0.61/hr (opens in new tab)	9 GB free · ~38 tok/s · $4.39 / 1M tokens
RTX A6000 (48 GB)17 GB free	~19 tok/s	$9.55	Deploy on Vast.aifrom $0.65/hr (opens in new tab)	48 GB · 17 GB free · ~19 tok/s · $9.55 / 1M tokens
RTX 6000 Ada (48 GB)17 GB free	~24 tok/s	$9.19	Deploy on Vast.aifrom $0.78/hr (opens in new tab)	48 GB · 17 GB free · ~24 tok/s · $9.19 / 1M tokens
L40S (48 GB)17 GB free	~21 tok/s	$10.45	Deploy on Vast.aifrom $0.80/hr (opens in new tab)	48 GB · 17 GB free · ~21 tok/s · $10.45 / 1M tokens
RTX PRO 6000 (96 GB)65 GB free	~39 tok/s	$11.95	Deploy on RunPodfrom $1.69/hr (opens in new tab)	96 GB · 65 GB free · ~39 tok/s · $11.95 / 1M tokens
H100 (80 GB)49 GB free	~83 tok/s	$9.06	Deploy on RunPodfrom $2.69/hr (opens in new tab)	80 GB · 49 GB free · ~83 tok/s · $9.06 / 1M tokens
H200 (141 GB)110 GB free	~118 tok/s	$8.44	Deploy on RunPodfrom $3.59/hr (opens in new tab)	141 GB · 110 GB free · ~118 tok/s · $8.44 / 1M tokens
B200 (192 GB)161 GB free	~197 tok/s	$8.43	Deploy on RunPodfrom $5.98/hr (opens in new tab)	192 GB · 161 GB free · ~197 tok/s · $8.43 / 1M tokens

Compare GPUs →