Fireworks runs AI models for businesses and helps customers customize downloadable, open-weight models using their own data. The company has worked with Uber, Shopify and Notion. Zhao said open models are now no more than six months behind frontier models, and probably closer to three months.
He said companies often start with leading proprietary models because they are familiar and low-risk. But once a product reaches 10,000 or 100,000 users and monthly token spending rises to $10,000 to $100,000, businesses should consider open models. Zhao expects frontier and open models to coexist: companies may use the most capable systems for complex planning while routing simpler tasks to cheaper alternatives.
Fireworks disclosed two security incidents this fall. In one, hackers accessed API keys and tokens stored as secrets for a limited number of accounts. In another, attackers breached its cloud provider, reached a development system and took customer data, including billing records and some datasets. Notion said AI chat transcripts from fewer than 1% of its workspaces were exposed and has disconnected from Fireworks. Uber said hackers stole personal information belonging to about 150,000 drivers through a vendor TechCrunch identified as Fireworks.
Zhao said the incidents affected a small subset of customers and that the company had no indication its promise of zero data retention for inference traffic had been compromised. He said Notion’s exposed data was uploaded through batch prediction, a service that stores datasets at rest, and that Fireworks is moving those services to an online flow.
Investors valued Fireworks at $17.5 billion in the summer. The company is reportedly in early talks about another funding round; Zhao declined to comment. The market is attracting other large investments: Together AI and Baseten have announced major rounds, while Palo Alto Networks said Wednesday it would buy Applied Compute for $4.5 billion in stock.

Comments (0)
No comments yet. Be the first!