Customers

Customer segments

• AI startups building AI-native products. They want cheap inference and the option to fine-tune or self-host as they scale.
• SaaS companies adding AI features. They want OpenAI-compatible APIs they can swap into without rewriting their integration.
• Enterprises with AI initiatives. They want open-source alternatives to closed-source frontier APIs for cost, customization, or data-control reasons.
• AI tooling platforms. Companies whose product is a layer over LLM APIs often use Together as one of their backends.
• Research customers. Academic or quasi-academic users running experiments on hosted models.

Why customers pick Together

• Per-token cost meaningfully below closed-source frontier APIs.
• OpenAI-API compatibility — minimal integration work to switch.
• Broad model catalog including the latest open-source releases quickly after their public availability.
• Lifecycle support — same platform for inference, fine-tuning, and training.
• Research-credibility-driven confidence in serving quality.
• Open posture on the underlying stack and models.

Who Together wins

• Cost-conscious AI builders looking for OpenAI alternatives.
• Customers fine-tuning models who want a managed fine-tuning experience.
• Customers who want one platform across the model lifecycle.
• Customers who value the research-credibility brand.

Who Together loses

• Customers who need absolute frontier capability — closed-source still wins on hardest tasks.
• Customers already deep into a hyperscaler ecosystem (AWS Bedrock, Azure AI, Vertex AI) who'd rather consolidate.
• Customers running production inference at scale enough to justify their own infrastructure (they'd be on CoreWeave or similar).
• Customers who need full control over the inference stack (they'd self-host on raw GPU).

Workload patterns

What customers actually run:

• Chatbots and conversational AI front-ends.
• RAG (retrieval-augmented generation) workflows.
• Document summarization and extraction.
• Code generation and assistance.
• Customer-support automation.
• Content generation (marketing, internal documents, etc.).
• Embedding generation for search and recommendation.
• Agentic workflows that string together multiple LLM calls.

The mix is broadly representative of the production-AI workload distribution across the industry.

Geographic distribution

Together's customer base is global. The cloud-API model travels well — customers anywhere with network access can use the platform. US and European customers dominate; Asia-Pacific is growing.

Takeaway

Together's customer base reflects the broader open-source-friendly slice of the AI market. The customer profile is distinct from CoreWeave's or Crusoe's — more SaaS, more startup, more cost-conscious. The next chapter examines competitive positioning.