Fireworks AI, an open-source model platform, is positioning itself as a cost-effective alternative to frontier AI labs like Anthropic and OpenAI. The company helps enterprises run smaller, cheaper open-weight models tuned on their own data, with customers including Uber, Shopify, and Notion.
Co-founder Chenyu Zhao said open models currently lag frontier models by at most six months, down from a year or more previously. He attributed slower adoption among businesses to brand familiarity and risk aversion before reaching scale, adding that companies spending $10,000 to $100,000 monthly on tokens typically consider switching.
Fireworks carries a $17.5 billion valuation from summer investors and is reportedly in early talks for another funding round, though Zhao declined to comment. Rival platforms Together AI and Baseten announced major rounds recently. Palo Alto Networks agreed to acquire Applied Compute, a similar startup, for $4.5 billion in stock.
The company disclosed two security incidents this fall. Hackers accessed API keys and tokens for a limited number of accounts in one breach, and breached a development system through a cloud provider in another. Notion disconnected from Fireworks after fewer than 1 percent of its workspaces had AI chat transcripts exposed. Uber said attackers stole names, addresses, birth dates, and driver's license numbers of roughly 150,000 drivers through what TechCrunch identified as Fireworks.
Zhao defended the company's security posture, stating Fireworks promises zero data retention for inference traffic with no indication that promise was broken. For batch prediction services where data is stored at rest, he said Fireworks is moving to an online flow model to minimize data retention.