Credits for venture backed AI startups, amount not published

How to claim it, in 4 steps

What the terms actually say

Fireworks publishes a detailed eligibility list and no number at all, which is the reverse of the usual shape and worth taking at face value rather than filling in from an aggregator. What is published is precise about scope: credits cover text and vision inference, image generation, speech to text, embeddings, model fine tuning and on-demand GPU deployments, and do not cover reserved compute commitments or professional services. The funding bar cuts from both ends, over $500k raised to qualify and no Series D or later, so this is aimed at companies with a fund behind them but not yet at scale. Credits expire one year after issue.

Who can claim it, in Fireworks AI's words

Founded within the last five years, a registered privately held for-profit company with a functional website, venture backed with over $500k raised, and not funded at Series D or later. The $500k floor rules out bootstrapped companies, which most programmes on this shelf do not.

What you get without this

Fireworks gives $1 in free credits on signup and then bills per token for serverless inference with postpaid billing and no setup fee. On-demand GPU deployments run from $7 to $8 per hour on H100 and H200 up to $20 per hour on GB300, and embeddings from $0.008 to $0.1 per million input tokens.

We list a perk because it beats walking up to the vendor unannounced. If it ever stops doing that, it comes off the index.