aquif-ai/aquif-3.5-Nano-1B
TEXT GENERATIONConcurrency Cost:1Model Size:2BQuant:BF16Ctx Length:32kLicense:apache-2.0Architecture:Transformer0.0K Open Weights Cold
aquif-ai/aquif-3.5-Nano-1B is a 1.72 billion parameter causal language model built on Qwen3, instruction-tuned for high performance-to-efficiency. It features a 40K token context window and bfloat16 precision, excelling in reasoning, mathematics, and code generation. This model is optimized for resource-constrained deployments and consumer-grade hardware, offering strong core capabilities across 10 languages.
Loading preview...
aquif-3.5-Nano-1B: Efficient and Capable Nano-Model
aquif-3.5-Nano-1B is a 1.72 billion parameter language model based on Qwen3, designed for strong performance in resource-constrained environments. It boasts a 40K token context window and bfloat16 precision, making it suitable for fast inference and minimal memory overhead on devices with 8GB+ VRAM.
Key Capabilities
- Efficient Architecture: Optimized layer configuration and extended context for complex reasoning.
- Strong Core Performance: Excels in multi-step logical inference, mathematical reasoning, and code generation.
- Multilingual Support: Native support for 10 languages including English, Chinese, and Japanese.
- Competitive Benchmarks: Achieves 72.9% on MMLU, 42.0% on GPQA Diamond, and 28.7% on AIME 2025, outperforming base Qwen3-1.7B and other comparable models.
Good For
- Edge Deployment: Real-time inference on resource-limited devices.
- API Services: Cost-effective inference at scale with low latency.
- Local Processing: Privacy-preserving on-device inference.
- Mathematical Reasoning & Code Assistance: Problem-solving and programming help within constrained budgets.