jhqxxx
4.9 KiB
4.9 KiB
Supported Models
aha supports a growing collection of state-of-the-art AI models across multiple domains.
Language Model
| Model | Parameters | Model Id | License |
|---|---|---|---|
| Qwen3-0.6B | 0.6B | Qwen/Qwen3-0.6B | Apache 2.0 |
| MiniCPM4-0.5B | 0.5B | OpenBMB/MiniCPM4-0.5B | Apache 2.0 |
| LFM2-1.2B | 1.2B | LiquidAI/LFM2-1.2B | lfm1.0 |
| LFM2.5-1.2B-Instruct | 1.2B | LiquidAI/LFM2.5-1.2B-Instruct | lfm1.0 |
Vision & Multimodal
| Model | Parameters | Model Id | License |
|---|---|---|---|
| Qwen2.5-VL | 3B 7B |
Qwen/Qwen2.5-VL-3B-Instruct Qwen/Qwen2.5-VL-7B-Instruct |
Qwen 研究许可协议 Apache 2.0 |
| Qwen3-VL | 2B 4B 8B 32B |
Qwen/Qwen3-VL-2B-Instruct Qwen/Qwen3-VL-4B-Instruct Qwen/Qwen3-VL-8B-Instruct Qwen/Qwen3-VL-32B-Instruct |
Apache 2.0 |
| Qwen3.5 | 0.8B 2B 4B 9B |
Qwen/Qwen3.5-0.8B Qwen/Qwen3.5-2B Qwen/Qwen3.5-4B Qwen/Qwen3.5-9B |
Apache 2.0 |
| LFM2.5-VL-1.6B | 1.6B | LiquidAI/LFM2.5-VL-1.6B | lfm1.0 |
| LFM2-VL-1.6B | 1.6B | LiquidAI/LFM2-VL-1.6B | lfm1.0 |
OCR
| Model | Languages | Model Id | License |
|---|---|---|---|
| PaddleOCR-VL | Multi | PaddlePaddle/PaddleOCR-VL PaddlePaddle/PaddleOCR-VL-1.5 |
Apache 2.0 |
| Hunyuan-OCR | Chinese | Tencent-Hunyuan/HunyuanOCR | Tencent Hunyuan Community License |
| DeepSeek-OCR | Multi | deepseek-ai/DeepSeek-OCR deepseek-ai/DeepSeek-OCR-2 |
MIT |
| GLM-OCR | 8 | ZhipuAI/GLM-OCR | MIT |
Speech Recognition (ASR)
| Model | Parameters | Language | Model Id | License |
|---|---|---|---|---|
| Fun-ASR-Nano-2512 | 2G | Chinese/English | FunAudioLLM/Fun-ASR-Nano-2512 | Not Specified |
| GLM-ASR-Nano-2512 | 4.5G | Chinese/English | ZhipuAI/GLM-ASR-Nano-2512 | MIT |
| Qwen3-ASR | 0.6B 1.7B |
Chinese/English | Qwen/Qwen3-ASR-0.6B Qwen/Qwen3-ASR-1.7B |
Apache 2.0 |
Audio Generation
| Model | version | Model Id | License |
|---|---|---|---|
| VoxCPM | 1 1.5 |
OpenBMB/VoxCPM-0.5B OpenBMB/VoxCPM1.5 |
Apache 2.0 |
Image Processing
| Model | Type | Model Id | License |
|---|---|---|---|
| RMBG-2.0 | Background Removal | AI-ModelScope/RMBG-2.0 | CC BY-NC 4.0 |
Model Sources
Models are sourced from:
- Hugging Face - Primary model hub
- ModelScope - Chinese model hub
Adding New Models
See Development Guide for instructions on adding new model integrations.
License Compliance
Important: Each model has its own license. Please review the model's license before use in production. Some key considerations:
- Apache 2.0: Permissive, commercial-friendly
- MIT: Permissive, commercial-friendly
- Qwen Research License: Research use, may have restrictions
- Tencent Hunyuan Community License: Custom license, review terms
- CC BY-NC 4.0: Non-commercial only
Always verify license terms before deployment in production environments.
Model Updates
Models updated from time to time.
Performance Benchmarks
Approximate inference speeds on CPU (M1 Pro):
| Model | Task | Tokens/sec |
|---|---|---|
| Qwen3-0.6B | Text | 40-50 |
| Qwen2.5-VL-3B | Vision | 20-30 |
| Qwen3-ASR-0.6B | ASR | 200-500x |
| VoxCPM-0.5B | TTS | Real-time |
Benchmarks vary by hardware and input size.