We're giving qualifying companies up to $20,000 in Arena training and inference credits to post-train a model on your own task and see the results for yourself. Fine-tune the latest open models or train an agent with reinforcement learning, then evaluate and deploy it, all in one place. We review every application and pick teams with a clear use case and data ready to go.
Most teams we speak to think an open model post-trained on their own data could match or beat a frontier API on their task, at a fraction of the cost. What holds them back is the upfront spend to find out, and the worry that a better model will land next month. We would rather cover that cost, so you can start now and run newer models through the same task as they come out.