Together AI Cloud (Serverless Inference, Fine-Tuning and GPU Clusters)

Offering overview

Serverless endpoints for open-weight models are the entry point: Llama, DeepSeek, Qwen, Kimi, GLM and gpt-oss families are served pay-per-token with cached-input discounts, batch inference and provisioned throughput with SLAs. Dedicated Model Inference and Dedicated Container Inference put the same stack on reserved hardware for custom models. Training features include supervised and DPO fine-tuning priced per million tokens, custom post-training at scale and an evaluations service. For raw capacity, GPU Clusters offer HGX H100, H200, B200, B300 and GB200/GB300 NVL72 systems, and AI Factory delivers custom infrastructure for the largest customers. A Sandbox product provides development environments for agentic workloads. Startups shipping AI products, research groups and enterprises standardising on open models all use the platform; the company is headquartered in San Francisco. Serverless pricing starts at $0.14 per million tokens (input and output) for Llama 3 8B Instruct Lite, Llama 3.3 70B is $1.04 per million, and dedicated H100 endpoints run $5.49 per GPU-hour. Cluster H100 capacity is listed at $3.99 per GPU-hour on-demand or $1.99 preemptible.

Technical specifications

Containers

Unknown

Kubernetes

Unknown

Root access

Unknown

API access

Unknown

Autoscaling

Unknown

Persistent processes

Unknown

AI capabilities

Model serving

Llama 3.3 70B, Llama 3 8B, DeepSeek V4, DeepSeek-V3.1, Qwen3.5, Qwen3.7, Qwen2.5, Kimi K3, Kimi K2.6, GLM-5.3, gpt-oss-120B

AI frameworks

Not stated

Inference endpoints

Unknown

Fine-tuning

Unknown

Vector database

Unknown

Model catalog

Unknown

Monitoring

Not stated

Evaluation

Not stated

Security & compliance

Geography

Provider headquarters are separate from offering infrastructure and availability.

Available plans

Pricing can change. Promotional prices are shown separately from standard prices.