Together AI

Developer Documentation & Integration

Together AI exposes an API for programmatic access to its inference, fine-tuning, and compute services. Developers can integrate with the platform using:

• PyTorch — native compatibility for ML workflows and model development
• CodeSandbox — for rapid prototyping and experimentation in browser-based environments

Inference options available via API:

• Serverless inference (per-token billing, no provisioning needed)
• Batch inference (up to 30 billion tokens per model)
• Provisioned throughput (reserved token-based capacity with SLAs)
• Dedicated model inference and dedicated container inference

Fine-tuning and model shaping are supported on the platform, enabling teams to adapt open-source base models (Llama, Qwen, DeepSeek, Gemma, GLM, MiniMax, Kimi, etc.) to domain-specific tasks.

GPU cluster access is available for teams requiring raw compute, with support for H100, H200, B200, GB200, and GB300 GPUs.

Pricing across all tiers is usage-based, with provisioned throughput and dedicated options offering SLA-backed guarantees for production workloads.