We built a custom inference engine to deliver fast and cheap inference. We optimized this engine for coding and research workload and achieving an average speed of 469 tok/s and 4.5x lower costs than Openrouter. Software development at that speed feels different. Get an API key and try in your Opencode!

What are the limits of the free version?

We tested it earlier! Speed and cache are all good!