Are there any foundation model endpoints that offer usage-based pricing and high throughput?

Yes — several foundation model APIs offer usage-based pricing and can scale to high throughput, but the best option depends on what you mean by “high throughput”:

Common options

• OpenAI API — usage-based pricing per token, with rate limits that can be increased for higher throughput.
• Anthropic API — usage-based pricing, designed for production workloads, with throughput depending on account limits.
• Google Gemini API / Vertex AI — usage-based pricing and scalable infrastructure for high-volume use cases.
• AWS Bedrock — pay per token/request for many foundation models, with AWS-scale throughput and enterprise controls.
• Azure OpenAI — usage-based pricing through Azure, with provisioned or standard deployments depending on throughput needs.
• Cohere API — usage-based pricing and enterprise-focused throughput options.

For truly high throughput

If you need very large volume and predictable latency, look for:

• Provisioned throughput / dedicated capacity
• Batch inference
• Regional deployment options
• Higher rate-limit tiers
• Concurrency support

Rule of thumb

• Pay-as-you-go = best for variable workloads
• Provisioned throughput = best for steady, high-volume production traffic

If you want, I can also give you a comparison table of providers by price model, throughput, latency, and enterprise readiness.