Chenyu Zhao, co-founder of Fireworks AI, has highlighted the opportunity for companies to use open-source AI models as an alternative to expensive frontier models. Speaking at the Slush startup conference in Helsinki, Zhao suggested that proprietary models are often unnecessary unless the latest generalised intelligence is required.
Fireworks AI runs AI models for businesses, focusing on open-weight models that can be downloaded and customised with a company's own data. This approach aims to make AI operations faster and cheaper than using models from companies like Anthropic and OpenAI.
The company has recently faced security challenges, disclosing two incidents this autumn. In one breach, hackers accessed API keys and tokens for a limited number of accounts. In another, attackers breached Fireworks AI's cloud provider, gaining access to a development system and customer data, including billing records and some datasets.
Following these incidents, Notion disconnected from Fireworks AI after chat transcripts from fewer than 1% of its workspaces were exposed. Uber also reported that hackers stole names, addresses, birth dates, and driver's license numbers of approximately 150,000 drivers through a vendor identified as Fireworks AI.
Zhao stated that open models are currently at most six months behind frontier models, and possibly closer to three months. He believes that while frontier models offer low risk for businesses seeking product-market fit, open-source alternatives become more appealing once monthly spending on tokens reaches significant levels.