LLM Fine-Tunes
Unlock the Power of Affordable, Open-Source LLMs, Matching the performance of GPT 3.5 at Half the Cost!
Ease of use - Abacus’s LLM fine-tunes are simple to create and easy to use. In fact, if you're already an OpenAI user, all you need to do is switch your API key
Performance - Pick a base model - Llama2 or Giraffe - and enable auto-fine-tune to reach performance on par with GPT 4.0
Inference Speed - Optimized for Fast Inference so your client APIs don’t have abnormal wait times
Fine-tune models using your data and/or get access to industry-specific and domain specific fine-tunes
Fine-tune Llama2 or Giraffe for your specific use case. Fine-tuning improves performance over the base model, and a well-executed fine-tune can match GPT-4’s performance and is 20 times cheaper. You can either upload a dataset and create a fine-tune or set up auto-fine-tuning to periodically fine-tune a base model using the requests sent to it.
Llama2 is a state-of-the-art open-source LLM launched by Meta. Giraffe is a context-extended, instruct fine-tune of Llama2 and can support 32k context length. Giraffe also achieves superior performance on various MT-Bench metrics such as reasoning and writing. More details on Giraffe can be found
here
