SiliconFlow — Fast Inference with Free Credits

SiliconFlow — high-performance LLM inference platform. Offers free credits to test their optimized kernels.

What's Included

  • Free credits on signup
  • 100+ models: Llama 3.1, Qwen 2.5, DeepSeek V2.5, Gemma 2, Phi-3.5, etc.
  • OpenAI-compatible API
  • Optimized kernels — FlashAttention, PagedAttention, etc.
  • High throughput — up to 10x faster than standard inference

Models

  • Llama 3.1 (8B, 70B, 405B)
  • Qwen 2.5 (7B, 32B, 72B)
  • DeepSeek V2.5 / V3
  • Gemma 2 (9B, 27B)
  • Phi-3.5 (Mini, Medium)
  • CodeLlama, CodeQwen, DeepSeek Coder

How to Start

  1. Sign up at siliconflow.cn
  2. Create API key
  3. Base URL: https://api.siliconflow.cn/v1

Bottom line: Great option for fast and cheap inference of open models.