Together AI

Listed at https://www.together.ai

Overall Rank #39 ⭐ Consider
❌ Proxy required (US-based service) | 🌍 International

💰 Token Pricing

TypePriceNote
Input $0.03-$1.74/MTok(模型相关) per million tokens
Output $0.12-$4.50/MTok(模型相关) per million tokens
Cache Read$0.06-$0.26/MTok(部分模型支持)Discounted
💡 Free Credits: $1 initial credit on signup (requires payment method)

🤖 Supported Models (200)

DeepSeek V4 ProQwen3.6-Plus / Qwen3.7-MaxQwen3.5 9B / Qwen3 235B-A22BMiniMax M3 / M2.5 / M2.7Kimi K2.7 Code / K2.6GLM-5.1 (Zhipu)Llama 3.3 70B TurboNVIDIA Nemotron 3 Ultra 550BCogito v2.1 671BGemma 4 31B / Gemma 3NFLUX.2 / FLUX.1 (Image gen)Imagen 4.0 / Seedream 4.0

✨ Pros

  • 200+ open-source and commercial models across chat/vision/image/video/audio/embedding
  • Fully OpenAI-compatible API — drop-in replacement for OpenAI SDK
  • FlashAttention-4 inference engine; 31% more TPS than TensorRT-LLM
  • Automatic cached input discounts; 50% off Batch API
  • Fine-tuning and GPU clusters (H100/B200/B300) available
  • ISO 27001:2022 certified, SOC 2 compliant

⚠️ Cons

  • ×No free tier ($1 initial credit only, requires payment method)
  • ×Not accessible from mainland China (US-based hosting)
  • ×New accounts have low default rate limits (dynamic scaling)

🎯 Best For

Developers seeking fast open-source model inference; AI agent & coding tool teams; production workloads needing OpenAI-compatible API