Have a large openAI GPT3.5 or 4 bill?
We can help automate the the distillation of your AI usage, into smaller cheaper RWKV models.
The process is outlined as the following
- Connect Recursal.AI router between your AI agents, and openAI backend
- Start lowering your cost, using our AI request cache, on the Recursal.AI router
- Wait for ~ 2 weeks as your dataset is being built up
- Finetune chosen datasets collected, into a 3B or 7B RWKV model
- Slowly reroute incrementally, your workload over to the finetuned model
- Watch your OpenAI bill drop down drastically (~ 50%)
The above flow allows you gain the best of both worlds, where you can offload common smaller tasks into finetuned efficient models. While still keeping new unique, or complicated tasks on the larger AI models
If you are interested in our pilot program for the above, fill up the form
