Fireworks

Fireworks (Fireworks AI, Inc.) provides an inference platform that serves open-source and fine-tuned models through serverless per-token pricing, on-demand GPU deployments and reserved capacity, plus managed training and reinforcement learning for custom models.

Customers are developers and enterprises building AI applications; the website names Cursor, Vercel, Notion, Quora, Sourcegraph and UiPath among users or partners. Revenue comes from usage-based token billing, GPU-second pricing for dedicated deployments, and training fees per million tokens.

The company is US-based, with job postings listing San Mateo, New York, San Francisco, London and Singapore. Its founding team is listed as former PyTorch and Meta engineers, and the careers page shows 88 open positions, mainly in engineering, sales, finance and legal.

Product: Serverless and dedicated inference for open-source models, managed fine-tuning and reinforcement learning, and the Fireworks Nexus model-routing product.