InternLMInternLMInternLM3ForCausalLM

Internlm3 8B Instruct — Hardware Requirements & GPU Compatibility

Chat

Internlm3 8B Instruct is a 8.8B-parameter open language model from InternLM in the InternLM family. It supports a context window of up to 32,768 tokens. At Q4_K_M it needs about 5.68 GB of VRAM — see which GPUs and Macs can run it below.

8.8B params · dense33K contextApache 2.0Jan 202557.7K downloads235 likes4.7K 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 8 GB (Q4_K_M · 32K)Best on 24 GB (Q4_K_M · ~115 tok/s · 32K)Macs from 8 GB (IQ2_M · ~21 tok/s)

Apache 2.0

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run Internlm3 8B Instruct

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 3060 12GB (5 GB headroom) · Q4_K_M

Generation speed: ~41 tok/s; generation speed
Cost per 1M output tokens: $0.56; per 1M output tokens

Deployon Vast.aiBest optionfrom $0.083/hr

Compare GPUs →

Run Internlm3 8B Instruct

6.4 GB VRAM at 16K context · fits the RTX 4090 (17.6 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/internlm3-8b-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=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/bartowski/internlm3-8b-instruct-GGUF, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get https://huggingface.co/bartowski/internlm3-8b-instruct-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 bartowski/internlm3-8b-instruct-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/bartowski/internlm3-8b-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/internlm3-8b-instruct-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 internlm/internlm3-8b-instruct --max-model-len 16384
Runs the original weights, not the Q4_K_M GGUF; needs ~19 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/internlm3-8b-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/internlm3-8b-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/internlm3-8b-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 3060 12GBCheapestCheapest5 GB free	~41 tok/s	$0.56	Deploy on Vast.aifrom $0.083/hr (opens in new tab)	5 GB free · ~41 tok/s · $0.56 / 1M tokens
RTX 3070 (8 GB)1 GB free	~51 tok/s	$0.70	Deploy on RunPodfrom $0.13/hr (opens in new tab)	8 GB · 1 GB free · ~51 tok/s · $0.70 / 1M tokens
RTX 3070 (8 GB)1 GB free	~51 tok/s	$0.73	Deploy on Vast.aifrom $0.14/hr (opens in new tab)	8 GB · 1 GB free · ~51 tok/s · $0.73 / 1M tokens
RTX 3060 Ti (8 GB)1 GB free	~51 tok/s	$0.81	Deploy on Vast.aifrom $0.15/hr (opens in new tab)	8 GB · 1 GB free · ~51 tok/s · $0.81 / 1M tokens
RTX 5060 Ti 16GB9 GB free	~51 tok/s	$0.89	Deploy on Vast.aifrom $0.16/hr (opens in new tab)	9 GB free · ~51 tok/s · $0.89 / 1M tokens
RTX 3080 Ti (12 GB)5 GB free	~104 tok/s	$0.48	Deploy on RunPodfrom $0.18/hr (opens in new tab)	12 GB · 5 GB free · ~104 tok/s · $0.48 / 1M tokens
Tesla V100 PCIe 16GB9 GB free	~103 tok/s	$0.51	Deploy on RunPodfrom $0.19/hr (opens in new tab)	9 GB free · ~103 tok/s · $0.51 / 1M tokens
RTX 4090 (24 GB)17 GB free	~115 tok/s	$0.82	Deploy on RunPodfrom $0.34/hr (opens in new tab)	24 GB · 17 GB free · ~115 tok/s · $0.82 / 1M tokens

Compare GPUs →

How Much VRAM Does Internlm3 8B Instruct Need?

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

VRAM	+32K	+128K	File	Quality	On RTX 4090 (reference)
2.8 GB	4.3 GB	—	2.6 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Fits: 32K context · ~232 tok/s
3.0 GB	4.6 GB	—	2.8 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Fits: 32K context · ~216 tok/s
3.1 GB	4.7 GB	—	3 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Fits: 32K context · ~208 tok/s
3.4 GB	4.9 GB	—	3.2 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Fits: 32K context · ~194 tok/s
3.8 GB	5.3 GB	—	3.6 GB	3-bit · i-quant — Importance-weighted 3-bit	Fits: 32K context · ~172 tok/s
3.9 GB	5.4 GB	—	3.2 GB	2-bit · small K-quant — 2-bit small K-quant	Fits: 32K context · ~167 tok/s
4.0 GB	5.5 GB	—	3.8 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Fits: 32K context · ~163 tok/s
4.1 GB	5.7 GB	—	3.5 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Fits: 32K context · ~158 tok/s
4.1 GB	5.7 GB	—	4 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Fits: 32K context · ~158 tok/s
4.3 GB	5.8 GB	—	4 GB	3-bit small — 3-bit small quantization	Fits: 32K context · ~154 tok/s
4.4 GB	5.9 GB	—	4.1 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Fits: 32K context · ~150 tok/s
4.7 GB	6.2 GB	—	4.4 GB	3-bit medium — 3-bit medium quantization	Fits: 32K context · ~140 tok/s
4.8 GB	6.3 GB	—	5.1 GB	4-bit · legacy — 4-bit legacy quantization	Fits: 32K context · ~137 tok/s
4.9 GB	6.4 GB	—	4.7 GB	3-bit large — 3-bit large quantization	Fits: 32K context · ~133 tok/s
5.1 GB	6.6 GB	—	4.8 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Fits: 32K context · ~128 tok/s
5.3 GB	6.9 GB	—	5.1 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Fits: 32K context · ~123 tok/s
5.3 GB	6.9 GB	—	5.6 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Fits: 32K context · ~123 tok/s
5.3 GB	6.9 GB	—	5.1 GB	4-bit small — 4-bit small quantization	Fits: 32K context · ~123 tok/s
5.7 GB	7.2 GB	—	5.4 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Fits: 32K context · ~115 tok/s
5.8 GB	7.3 GB	—	5.7 GB	4-bit large — 4-bit large quantization	Fits: 32K context · ~113 tok/s
5.9 GB	7.4 GB	—	6.1 GB	5-bit · legacy — 5-bit legacy quantization	Fits: 32K context · ~111 tok/s
6.5 GB	8 GB	—	6.1 GB	5-bit small — 5-bit small quantization	Fits: 32K context · ~102 tok/s
6.7 GB	8.2 GB	—	6.3 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Fits: 32K context · ~98 tok/s
6.8 GB	8.3 GB	—	6.6 GB	5-bit large — 5-bit large quantization	Fits: 32K context · ~97 tok/s
7.7 GB	9.2 GB	—	7.2 GB	6-bit, very good — 6-bit quantization, very good quality	Fits: 32K context · ~86 tok/s
9.2 GB	10.7 GB	—	9.4 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Fits: 32K context · ~71 tok/s
18.0 GB	19.5 GB	—	~17.6 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Fits: 32K context · ~36 tok/s

est.calculated, no published file yet

Selected Q4_K_M

4-bit medium quantization

6.4GBat 16K context

Weights: 5.3 GB
KV cache: 0.8 GB
Overhead: 0.3 GB

Fits RTX 4090 (reference) · 17.6 GB free · ~115 tok/s

≥50 tok/s needs ≥436 GB/s · RTX 5090 ~205, RTX 3090 Ti ~115

Which Hardware Can Run Internlm3 8B Instruct?

Q4_K_M · 5.7 GB

Internlm3 8B Instruct (Q4_K_M) requires 5.7 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 8+ GB is recommended. Using the full 33K context window can add up to 1.5 GB, bringing total usage to 7.2 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3070 Ti. 151 devices with unified memory can run Internlm3 8B 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 · 6.4 GB · ~31 tok/s · 32K6.4~3132K8 GB tier →

12 GB · RTX 3060 12GBQ4_K_M · 6.4 GB · ~41 tok/s · 32K6.4~4132K12 GB tier →

16 GB · RTX 5060 Ti 16GBQ4_K_M · 6.4 GB · ~51 tok/s · 32K6.4~5132K16 GB tier →

24 GB · RTX 4090referenceQ4_K_M · 6.4 GB · ~115 tok/s · 32K6.4~11532K24 GB tier →

Mac 32 GB · Mac Mini M4Q4_K_M · 6.4 GB · ~15 tok/s · 32K6.4~1532KMacs →

Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 6.4 GB · ~67 tok/s · 32K6.4~6732KMacs →

How good is Internlm3 8B 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-Pro
TIGER-LabIndependent
57.6%
#61/145 open
Independent

Internlm3 8B Instruct benchmark results

Alternatives to Internlm3 8B Instruct

InternLM family · 7 models →

Similar models

6tok/s on RTX 4090

Model	Score	VRAM	tok/s	Why it is similar
Granite Switch 4.1 8BIBM	—	20.7 GB	~36	same size
ALoDLM 8BAmazon	—	7.6 GB	~119	same size
Llm Jp 4.1 8B Thinkingllm-jp	—	7.6 GB	~115	same size
Apertus V1.5 8Bswiss-ai	—	8.5 GB	~111	same size
Ornith 1.0 9BDeep Reinforce	—	6.5 GB	~106	same size
LFM2.5 8B A1BLiquid AI	—	5.6 GB	~335	same size

Specifications

Publisher: InternLM
Family: InternLM
Parameters: 8.8B
Architecture: InternLM3ForCausalLM
Context Length: 32,768 tokens
Vocabulary Size: 128,512
Release Date: 2025-01-13
License: Apache 2.0

Resources

Frequently Asked Questions

How much VRAM does Internlm3 8B Instruct need?

Internlm3 8B Instruct requires 5.7 GB of VRAM at Q4_K_M, or 18.0 GB at BF16. Full 33K context adds up to 1.5 GB (7.2 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 8.8B × 4.8 bits ÷ 8 = 5.3 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 ≈ 1.9 GB (at full 33K context)

VRAM usage by quantization

Q4_K_M

5.7 GB

Q4_K_M + full context

7.2 GB

Learn more about VRAM estimation →

What's the best quantization for Internlm3 8B Instruct?

For Internlm3 8B Instruct, Q4_K_M (5.7 GB) offers the best balance of quality and VRAM usage. Q4_K_L (5.8 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 2.8 GB.

VRAM requirement by quantization

IQ2_XXS

2.8 GB~53%

Q2_K

4.1 GB~75%

Q3_K_L

4.9 GB~86%

Q4_K_M ★

5.7 GB~89%

Q5_0

5.9 GB~90%

BF16

18.0 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Internlm3 8B Instruct on a Mac?

Yes — Mac Mini M2 (8 GB) and 145 other Macs can run Internlm3 8B 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 5.7 GB of usable unified memory (RAM minus macOS overhead), from Mac Mini M2 Pro (16 GB).

Can I run Internlm3 8B Instruct locally?

Yes — Internlm3 8B Instruct can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Internlm3 8B Instruct?

At Q4_K_M, Internlm3 8B Instruct can reach ~845 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~115 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 ÷ 5.6800 × 0.65 = ~915 tok/s

Estimated speed at Q4_K_M (5.7 GB)

NVIDIA B200

~916 tok/s

NVIDIA GeForce RTX 4090

~115 tok/s

NVIDIA B300

~916 tok/s

AMD Instinct MI350X

~845 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 Internlm3 8B Instruct?

At Q4_K_M, the download is about 5.28 GB. The full-precision BF16 version is 17.61 GB. The smallest option (IQ2_XXS) is 2.42 GB.

Which GPUs can run Internlm3 8B Instruct?

52 consumer GPUs can run Internlm3 8B 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. 40 GPUs have plenty of headroom for comfortable inference.

Which devices can run Internlm3 8B Instruct?

151 devices with unified memory can run Internlm3 8B Instruct at Q4_K_M (5.7 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.