Qwen2.5 Omni 3B MNN — Hardware Requirements & GPU Compatibility
Chat
Qwen2.5 Omni 3B MNN is a 3B-parameter open language model from taobao-mnn in the Qwen 2.5 family. At BF16 it needs about 6.60 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 8 GB (BF16 · 8K)Best on 24 GB (BF16 · ~99 tok/s · 32K)Macs from 16 GB (BF16 · ~21 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Qwen2.5 Omni 3B MNN
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve taobao-mnn/Qwen2.5-Omni-3B-MNN --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on RTX 3060 12GB (3 GB headroom) · BF16
- Generation speed
- ~36 tok/s
- generation speed
- Cost per 1M output tokens
- $0.54
- per 1M output tokens
How Much VRAM Does Qwen2.5 Omni 3B MNN Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 6.6 GB | 10.8 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 32K context · ~99 tok/s |
Selected BF16
Brain floating point 168.6GBat 16K context
- Weights
- 6.0 GB
- KV cache
- 2.2 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 15.5 GB free · ~99 tok/s
≥50 tok/s needs ≥507 GB/s · RTX 5090 ~177, RTX 3090 Ti ~99
Which Hardware Can Run Qwen2.5 Omni 3B MNN?
BF16 · 6.6 GBQwen2.5 Omni 3B MNN (BF16) requires 6.6 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 9+ GB is recommended. 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 Qwen2.5 Omni 3B MNN, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), NVIDIA Jetson Orin Nano 8GB (Super).
12 GB
12 GB tier →fits · 3.5 GB free · up to 32K ctx
10 GPUs · 1 device family
RTX 3080 Ti~90Simulate output on RTX 3080 TiRTX 5070~66Simulate output on RTX 5070RTX 4070~50Simulate output on RTX 4070iPhone~8Simulate output on Apple iPhone 17 Pro20 GB
fits · 11.4 GB free · up to 32K ctx
1 GPU
RX 7900 XT~73Simulate output on RX 7900 XT32–40 GB
fits · 23.4 GB free · up to 32K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~177Simulate output on RTX 5090M5 Max36 GB~49Simulate output on Mac Studio M5 Max (36 GB)M532 GB~16Simulate output on MacBook Air 13" M5 (32 GB)Orin~20Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~13Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at BF16
Alternatives to Qwen2.5 Omni 3B MNN
Qwen 2.5 family · 43 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| D1 3BLiquid AI | — | 2.6 GB | ~285 | same size |
| SmolLM3 3BHugging Face | — | 3.4 GB | ~285 | same size |
| Llama 3.2 3B InstructMeta | — | 4.1 GB | ~309 | same size |
| Phi 2Microsoft | — | 2.6 GB | ~248 | same size |
| AI21 Jamba Reasoning 3BAI21 Labs | — | 2.3 GB | ~291 | same size |
| Quasar 3B A1B Previewsilx-ai | — | 8.4 GB | ~167 | same size |
Specifications
- Publisher
- taobao-mnn
- Family
- Qwen 2.5
- Parameters
- 3B
- Release Date
- 2025-05-07
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Qwen2.5 Omni 3B MNN need?
Qwen2.5 Omni 3B MNN requires 6.6 GB of VRAM at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 3B × 16 bits ÷ 8 = 6 GB
KV Cache + Overhead ≈ 0.6 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
Can I run Qwen2.5 Omni 3B MNN on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run Qwen2.5 Omni 3B MNN. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 6.6 GB of usable unified memory (RAM minus macOS overhead).
Can I run Qwen2.5 Omni 3B MNN locally?
Yes — Qwen2.5 Omni 3B MNN can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Qwen2.5 Omni 3B MNN?
At BF16, Qwen2.5 Omni 3B MNN can reach ~727 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~99 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 ÷ 6.6 × 0.65 = ~788 tok/s
Estimated speed at BF16 (6.6 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Qwen2.5 Omni 3B MNN?
At BF16, the download is about 6.00 GB.
Which GPUs can run Qwen2.5 Omni 3B MNN?
52 consumer GPUs can run Qwen2.5 Omni 3B MNN at BF16. 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 Qwen2.5 Omni 3B MNN?
151 devices with unified memory can run Qwen2.5 Omni 3B MNN at BF16 (6.6 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.