Together AI pricing: per-model API rates
Together AI sells serverless inference (pay per 1M tokens), dedicated GPUs, fine-tuning and clusters. There is no simple monthly Pro seat. Samples below are from together.ai/pricing on 2026-11-06 — always re-check the live model table.
Llama 3 8B Instruct Lite$0.14 / $0.14Input / output per 1M tokens
gpt-oss-120B$0.15 / $0.60Input / output per 1M tokens
GLM-5.3-Flash$0.15 / $0.50Cached input $0.03
Dedicated H100$5.49 / GPU / hrOn-demand dedicated inference
All Together AI plans
Per-seat prices in USD from the official pricing pages. “Annual billing” is the per-month price when billed annually, as the vendor states it; the yearly total is shown where published.
Swipe sideways to see all columns
| Plan | Monthly billing | Annual billing | Seats | Usage (as published) |
|---|
Sources: .
Sample serverless rates (USD per 1M tokens)
| Feature | Input | Output | Notes |
|---|---|---|---|
| Llama 3 8B Instruct Lite | $0.14 | $0.14 | Chat |
| gpt-oss-120B | $0.15 | $0.60 | Chat |
| GLM-5.3-Flash | $0.15 (cached $0.03) | $0.50 | Chat |
| DeepSeek V4.1 Flash | $0.30 (cached $0.006) | $1.20 | Chat |
| Kimi K3 | $3.00 (cached $0.30) | $15.00 | Chat |
| Dedicated NVIDIA HGX H100 | $5.49/GPU/hr | — | On-demand dedicated |
| GPU cluster H100 on-demand | $3.99/GPU/hr | — | Clusters table |
- Whisper Large v3 transcription sample: $0.0015/audio minute on the same page.
Source: https://www.together.ai/pricing (checked 2026-11-06). Full catalog is much larger — confirm the exact model ID before budgeting.
How Together AI pricing works
Serverless vs dedicated
Serverless is pay-per-token across a large open-model catalog. Dedicated and cluster SKUs bill per GPU-hour for reserved capacity.
How we check prices
Official Together AI pricing page. Corrections: [email protected].
Related
- Compare Together AI with other AI subscriptions by team size
- API or subscription? Break-even calculator
- LLM API workload cost calculator
- ChatGPT pricing
- Claude pricing
- Cursor pricing
- Grok pricing
- GitHub Copilot pricing
- Gemini pricing
- Perplexity pricing
- Microsoft 365 Copilot pricing
- Windsurf pricing
- Midjourney pricing
- ElevenLabs pricing
- Replit pricing
- Lovable pricing
- Canva AI pricing
- Notion AI pricing
- Grammarly pricing
- Runway pricing
- HeyGen pricing
- Kling AI pricing
- DeepSeek pricing
- Suno pricing
- Bolt.new pricing
- v0 pricing
- Zapier pricing
- n8n pricing
- Make pricing
- Gamma pricing
- Otter.ai pricing
- Fireflies.ai pricing
- Jasper pricing
- Synthesia pricing
- Descript pricing
- Leonardo.Ai pricing
- Adobe Firefly pricing
- Microsoft Copilot pricing
- Mistral Le Chat pricing
- Poe pricing
- OpenRouter pricing
- Claude Code pricing
- OpenAI Codex pricing
- Kiro pricing
- xAI API pricing
- OpenAI API pricing
- Anthropic API pricing
- CapCut pricing
- Intercom Fin pricing
- Pika pricing
- Zendesk AI pricing
- Flux pricing
- Luma Dream Machine pricing
- Qwen pricing
- Ideogram pricing
- Ollama pricing
- Cline pricing
- Google Workspace with Gemini pricing
- Semrush AI pricing
- Copy.ai pricing
- Writesonic pricing
- Surfer SEO pricing
- tl;dv pricing
- Firecrawl pricing
- Sudowrite pricing
- CodeRabbit pricing
- Langfuse pricing
- Krisp pricing
- Frase pricing
- Clearscope pricing
- LangSmith pricing
- Devin pricing
- Granola pricing
- Amazon Q Developer pricing
- Read.ai pricing
- Rytr pricing
- Mem pricing
- Amp (Sourcegraph) pricing
- You.com pricing
- Wordtune pricing
- Zoom AI Companion (ZoomMate) pricing
- Recraft pricing
- Stability AI pricing
- Zed AI pricing
- LM Studio pricing
- Hailuo AI pricing
- Slack AI pricing
- ClickUp Brain pricing
- Photoroom pricing
- Hugging Face pricing
- Character.AI pricing
- JetBrains AI pricing
- Lindy pricing
Together AI pricing FAQ
How much is Together AI?
Usage-based. Example: Llama 3 8B Instruct Lite $0.14/$0.14 per 1M tokens; dedicated H100 $5.49/GPU/hr in our capture.
Is there a monthly seat?
No public Pro-style seat on the pricing page we captured — you pay for tokens and hardware.
Is Together AI billed per token or per GPU?
Both. Serverless inference is pay per 1M tokens; dedicated and cluster SKUs bill per GPU-hour for reserved capacity.
Do cached tokens cost less?
On models that list a cached rate, yes. For example GLM-5.3-Flash input is $0.15 per 1M tokens, or $0.03 when cached.
Where do I confirm the exact model price?
On together.ai/pricing. The model catalog is much larger than our samples, so confirm the exact model ID before budgeting.