InternVL3 38B on NVIDIA RTX 2000 Ada Generation

Can the NVIDIA RTX 2000 Ada Generation run InternVL3 38B?

Yes, but slowly (offloads to RAM)

Slow (offloads to RAM). At Q2_K it needs about 18.9 GB of 16.0 GB usable, with an estimated speed of 8.1 tok/s.

Best quant: Q2_K
Generation speed: 8.1 tok/s
Prompt processing: 195 tok/s
First token (1K prompt): 5.2 s
Largest safe context: —

Best app to run it on the NVIDIA RTX 2000 Ada Generation

On Windows

LM Studio

Desktop app · runs on llama.cpp, MLX

Splits the model between GPU memory and system RAM, which this setup needs.
Point-and-click: search, download and chat in one window.

Open InternVL3 38B in LM Studio

llama.cpp Splits the model between GPU memory and system RAM, which this setup needs.
Ollama Splits the model between GPU memory and system RAM, which this setup needs.

On Linux

LM Studio

Desktop app · runs on llama.cpp, MLX

Splits the model between GPU memory and system RAM, which this setup needs.
Point-and-click: search, download and chat in one window.

Open InternVL3 38B in LM Studio

llama.cpp Splits the model between GPU memory and system RAM, which this setup needs.
Ollama Splits the model between GPU memory and system RAM, which this setup needs.

How we choose: the files published for this model, your operating system and GPU, whether it fits fully in memory, and published speed tests (MLX vs llama.cpp on Apple silicon, vLLM vs llama.cpp vs Ollama on NVIDIA). Details.

Download InternVL3 38B

Recommended for the NVIDIA RTX 2000 Ada Generation: Q2_K. Every other quantization has its own download link in the table below.

Quantized files · GGUF

Works in LM Studio, llama.cpp, Jan and Ollama. No sign-in needed.

Download the Q2_K file (opens in a new tab)

huggingface.co/unsloth/InternVL3-38B-GGUF (opens in a new tab)

ollama run hf.co/unsloth/InternVL3-38B-GGUF:Q2_K

Original weights · Hugging Face

Full-precision files from Shanghai AI Lab, for fine-tuning, vLLM, Transformers.

huggingface.co/OpenGVLab/InternVL3-38B (opens in a new tab)

No sign-in needed.

Links checked 2 days ago against Hugging Face, Ollama and GitHub.

How it’s rated

Quality 49Fair quality (C) · 49 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 53/100 (quality, fit, speed and memory use) sets the ranking.: How capable the model itself is at Q2_K, from its editorial quality tier scaled by the quantization or precision. It’s the same on any hardware that runs this variant. Scale: A+ 85+ · A 70+ · B 55+ · C 40+ · D under 40.
Runs Slowoffloads to RAM: How well it runs on the NVIDIA RTX 2000 Ada Generation: speed, capped by how well it fits in memory. Runs great: 30+ tokens/s (or near the image, video or decision speed target) · Runs well: 10+ tokens/s · Slow: 3+ tokens/s · Very slow: under 3 · Won’t run: doesn’t fit even with offload. A tight fit or RAM offload caps it at Runs well; heavy offload caps it at Slow.

Match score

To order lists, each result also gets a match score out of 100 from four parts. It favors the most capable model that still runs well.

Model qualityEditorial quality tier of the model, scaled by the quantization or precision used.	22.1 / 45
Memory fitFull marks when everything sits in GPU memory; fewer when part spills to system RAM.	10.0 / 25
SpeedGeneration speed against a 40 tokens/s target for LLMs, or render time against a target for image and video.	6.8 / 15
Memory useRewards using the memory you have, so a bigger, better variant beats a tiny one that leaves most memory idle.	15.0 / 15
Match score	53.8 / 100

The match score orders results; it is not shown as a grade. Quality and Runs are the two ratings in the tables.

Every quantization on the NVIDIA RTX 2000 Ada Generation

Quant	Memory needed	Available	Speed	Runs	Quality	Download
`Q2_K`	18.9 GB	16.0 GB	8.1 tok/s	Slowoffloads to RAM	49Fair quality (C) · 49 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 53/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q3_K_M`	22.4 GB	16.0 GB	5.2 tok/s	Slowheavy RAM offload	58Good quality (B) · 58 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 50/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q4_K_M`	27.0 GB	16.0 GB	3.4 tok/s	Slowheavy RAM offload	66Good quality (B) · 66 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 53/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q5_K_M`	30.8 GB	16.0 GB	2.6 tok/s	Very slowheavy RAM offload	67Good quality (B) · 67 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 53/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q6_K`	35.0 GB	16.0 GB	2.0 tok/s	Very slowheavy RAM offload	69Good quality (B) · 69 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 53/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q8_0`	44.2 GB	16.0 GB	—	Won't run	70Great quality (A) · 70 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 9/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (UD-Q8_K_XL) (opens in a new tab)
`F16`	80.3 GB	16.0 GB	—	Won't run	70Great quality (A) · 70 of 100: how capable the model is at F16, whatever it runs on. Match score 9/100 (quality, fit, speed and memory use) sets the ranking.	GGUF · folder (opens in a new tab) Original weights (opens in a new tab)

Memory figures assume 8K of context. Longer conversations or documents grow the KV cache.

Other hardware that runs it

67Good quality (B) · 67 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 84/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 5090 Q5_K_M · 47.3 tok/s · Runs well (tight fit)
69Good quality (B) · 69 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 80/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA RTX PRO 6000 Blackwell Q6_K · 39.8 tok/s · Runs great
58Good quality (B) · 58 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 80/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 4090 Q3_K_M · 41.7 tok/s · Runs well (tight fit)
58Good quality (B) · 58 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 80/100 (quality, fit, speed and memory use) sets the ranking. NVIDIA GeForce RTX 3090 Q3_K_M · 38.7 tok/s · Runs well (tight fit)
58Good quality (B) · 58 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 79/100 (quality, fit, speed and memory use) sets the ranking. AMD Radeon RX 7900 XTX Q3_K_M · 34.2 tok/s · Runs well (tight fit)
66Good quality (B) · 66 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 78/100 (quality, fit, speed and memory use) sets the ranking. Apple M4 Max (64 GB) Q4_K_M · 15.0 tok/s · Runs well

Quick answers

How much memory does InternVL3 38B need on the NVIDIA RTX 2000 Ada Generation 16GB?

At Q2_K it needs about 18.9 GB. The NVIDIA RTX 2000 Ada Generation 16GB has 16.0 GB usable.

How fast is InternVL3 38B on the NVIDIA RTX 2000 Ada Generation 16GB?

About 8.1 tok/s at Q2_K, an estimate from the hardware’s memory bandwidth and the size of the model.

Which quantization of InternVL3 38B should I use on the NVIDIA RTX 2000 Ada Generation 16GB?

Q2_K. Of the versions that fit, it is the one that runs best; when two run equally well, the check prefers the higher-rated one.

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.