Best Local AI Models for NVIDIA RTX 4090 (24 GB VRAM)
The NVIDIA RTX 4090 has 24 GB of VRAM. Here are the popular AI models it can run locally (4,096-token context, ~32.0 GB system RAM assumed), ranked by popularity.
See also: Best GPU for running local LLMs.
The NVIDIA RTX 4090 is built on NVIDIA's Ada Lovelace architecture featuring GDDR6X 384-bit delivering 1008 GB/s of raw memory bandwidth. The flagship 24 GB gold standard: over 1 TB/s bandwidth delivers blistering 100+ tokens/s on 8B–14B models.
Equipped with 24 GB of dedicated VRAM, the NVIDIA RTX 4090 can run 36 popular open-source models completely in GPU memory without offloading. This includes full-speed execution for weights like Qwen3-Coder-30B-A3B-Instruct-GGUF, LFM2.5-2.6B-GGUF, LFM2.5-8B-A1B-GGUF. For a comprehensive breakdown of compatible model weights, see our guide to the best LLMs for 24 GB VRAM.
With a memory bandwidth of 1008 GB/s, this card can generate tokens at an estimated peak rate of ~168.0 tokens/second on an 8B parameter model (Q4_K_M). Its rated power draw is 450W TDP, so ensure your system's power supply and case ventilation are adequate for sustained local inferencing.
36 fit fully in VRAM · 1 run with offload
| Model | Size | Quant. | Quality | Memory | Speed~ | Verdict |
|---|---|---|---|---|---|---|
| unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF | 30.53B | Q5_K_XL | Very good |
21.42 GB
|
19.8 t/s | Fits in VRAM |
| LiquidAI/LFM2.5-2.6B-GGUF | 2.7B | BF16 | Excellent |
5.89 GB
|
79.5 t/s | Fits in VRAM |
| LiquidAI/LFM2.5-8B-A1B-GGUF | 8.47B | BF16 | Excellent |
16.63 GB
|
25.3 t/s | Fits in VRAM |
| LiquidAI/LFM2.5-230M-GGUF | 0.23B | BF16 | Excellent |
1.28 GB
|
929.9 t/s | Fits in VRAM |
| unsloth/gpt-oss-20b-GGUF | 20.91B | F16 | Very good |
13.74 GB
|
31.1 t/s | Fits in VRAM |
| Qwen/Qwen3-8B-GGUF | 8.19B | Q8_0 | Excellent |
9.47 GB
|
49.3 t/s | Fits in VRAM |
| unsloth/Ornith-1.0-9B-GGUF | — | BF16 | Excellent |
17.61 GB
|
24.0 t/s | Fits in VRAM |
| bartowski/DeepSeek-Coder-V2-Lite-Instruct-GGUF | 15.71B | Q8_0_L | Excellent |
17.39 GB
|
25.1 t/s | Fits in VRAM |
| unsloth/Qwen3-4B-GGUF | 4.02B | BF16 | Excellent |
8.86 GB
|
53.3 t/s | Fits in VRAM |
| unsloth/Qwen-AgentWorld-35B-A3B-GGUF | 34.66B | Q4_K_XL | Very good |
21.67 GB
|
19.2 t/s | Fits in VRAM |
| bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF | 34.66B | Q5_K_S | Very good |
23.38 GB
|
17.8 t/s | Fits in VRAM |
| Qwen/Qwen3-0.6B-GGUF | 0.75B | Q8_0 | Excellent |
1.83 GB
|
671.7 t/s | Fits in VRAM |
| bartowski/Meta-Llama-3.1-8B-Instruct-GGUF | 8.03B | Q8_0 | Excellent |
9.23 GB
|
50.3 t/s | Fits in VRAM |
| LiquidAI/LFM2.5-1.2B-Instruct-GGUF | 1.17B | BF16 | Excellent |
3.03 GB
|
183.3 t/s | Fits in VRAM |
| Qwen/Qwen2.5-Coder-7B-Instruct-GGUF | 7.62B | GGUF | Excellent |
15.21 GB
|
28.2 t/s | Fits in VRAM |
| janhq/Jan-v3.5-4B-gguf | 4.41B | GGUF | Excellent |
9.59 GB
|
48.6 t/s | Fits in VRAM |
| MaziyarPanahi/Qwen3-14B-GGUF | 14.77B | Q6_K | Excellent |
12.71 GB
|
35.4 t/s | Fits in VRAM |
| MaziyarPanahi/Qwen3-1.7B-GGUF | 2.03B | GGUF | Excellent |
5.03 GB
|
105.5 t/s | Fits in VRAM |
| MaziyarPanahi/Qwen3-30B-A3B-GGUF | 30.53B | Q5_K_M | Very good |
21.41 GB
|
19.8 t/s | Fits in VRAM |
| MaziyarPanahi/Qwen3-32B-GGUF | 32.76B | Q5_K_M | Very good |
23.42 GB
|
18.5 t/s | Fits in VRAM |
| ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF | 31.58B | Q4_0 | Good |
18.42 GB
|
22.7 t/s | Fits in VRAM |
| Qwen/Qwen2.5-1.5B-Instruct-GGUF | 1.54B | GGUF | Excellent |
4.23 GB
|
120.6 t/s | Fits in VRAM |
| MaziyarPanahi/Yi-Coder-9B-Chat-GGUF | 8.83B | GGUF | Excellent |
17.62 GB
|
24.3 t/s | Fits in VRAM |
| bartowski/Qwen2.5-32B-Instruct-GGUF | 32.76B | Q5_K_L | Very good |
23.91 GB
|
18.1 t/s | Fits in VRAM |
| MaziyarPanahi/Qwen3-4B-Instruct-2507-GGUF | 4.02B | GGUF | Excellent |
8.86 GB
|
53.3 t/s | Fits in VRAM |
| Qwen/Qwen2.5-0.5B-Instruct-GGUF | 0.49B | GGUF | Excellent |
2.03 GB
|
339.1 t/s | Fits in VRAM |
| Qwen/Qwen2.5-3B-Instruct-GGUF | 3.09B | GGUF | Excellent |
7.27 GB
|
63.2 t/s | Fits in VRAM |
| bartowski/Qwen2.5-7B-Instruct-GGUF | 7.62B | F16 | Excellent |
15.21 GB
|
28.2 t/s | Fits in VRAM |
| unsloth/Llama-3.2-3B-Instruct-GGUF | 3.21B | F16 | Excellent |
7.23 GB
|
66.8 t/s | Fits in VRAM |
| MaziyarPanahi/Meta-Llama-3-8B-Instruct-GGUF | 8.03B | GGUF | Excellent |
16.24 GB
|
26.7 t/s | Fits in VRAM |
| unsloth/Qwen3-Coder-Next-GGUF | 79.67B | IQ2_XXS | Very low |
22.89 GB
|
18.4 t/s | Fits in VRAM |
| MaziyarPanahi/Yi-Coder-1.5B-Chat-GGUF | 1.48B | GGUF | Excellent |
4.3 GB
|
145.4 t/s | Fits in VRAM |
| MaziyarPanahi/Phi-3.5-mini-instruct-GGUF | 3.82B | Q8_0 | Excellent |
6.08 GB
|
105.8 t/s | Fits in VRAM |
| MaziyarPanahi/Mistral-7B-Instruct-v0.3-GGUF | 7.25B | GGUF | Excellent |
14.8 GB
|
29.6 t/s | Fits in VRAM |
| MaziyarPanahi/gemma-3-4b-it-GGUF | 4.3B | GGUF | Excellent |
8.38 GB
|
55.3 t/s | Fits in VRAM |
| MaziyarPanahi/Llama-3.2-1B-Instruct-GGUF | 1.24B | GGUF | Excellent |
3.3 GB
|
173.2 t/s | Fits in VRAM |
| MaziyarPanahi/Mixtral-8x22B-v0.1-GGUF | 140.62B | IQ3_XS | Fair |
55.9 GB
|
0.9 t/s | Offload |
"Fits in VRAM" = fast, fully on GPU. "Offload" = part on system RAM, slower. Speed is a rough estimate.
Frequently asked questions
What is the VRAM and memory bandwidth of the NVIDIA RTX 4090?
The NVIDIA RTX 4090 features 24 GB of VRAM and a memory bandwidth of 1008 GB/s (GDDR6X 384-bit). In local language model inference, VRAM determines which model sizes fit on the card, while memory bandwidth dictates how many tokens per second the GPU generates.
What is the best local AI model to run on a NVIDIA RTX 4090?
The best overall model for the NVIDIA RTX 4090 is unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF using the recommended Q5_K_XL quantization (21.42 GB total memory). With 24 GB of VRAM, this GPU typically runs a 30–35B model at Q5, fully on the GPU at full GPU speed. Check our guide to the best LLMs for 24 GB VRAM for details.
Can the NVIDIA RTX 4090 run 70B parameter models like Llama 3.3 70B?
Yes. With 24 GB of VRAM, the NVIDIA RTX 4090 can run 70B parameter models quantized to 3-bit or 4-bit (such as IQ3_M or Q4_K_M) with moderate context windows (4k tokens). For larger context windows (16k+), pairing with a second GPU or slight CPU offloading may be required.
What power supply (PSU) do you need for the NVIDIA RTX 4090 when running local AI?
The NVIDIA RTX 4090 has a rated TDP of 450W. While local inference typically draws less power than full 3D rasterization or gaming, continuous token generation can sustain high loads. A quality power supply of at least 750W is strongly recommended.
NVIDIA RTX 4090 Head-to-Head Comparisons
Compare specs, memory bandwidth, and AI model capability against other graphics cards.
RTX 3090 24GB vs RTX 4090 24GB
Used value king vs new flagship for 32B and 70B models.
RTX 5090 32GB vs RTX 4090 24GB
32 GB GDDR7 vs 24 GB: what new models become possible.
RTX 4080 Super 16GB vs RTX 4090 24GB
16 GB vs 24 GB: the threshold for 32B models.
AMD RX 7900 XTX vs NVIDIA RTX 4090
24 GB VRAM faceoff: ROCm vs CUDA for local AI.