lm-sys / FastChat

An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena.
Apache License 2.0
35.53k stars 4.36k forks source link

[Chatbot Arena] Add GLM-4 variants: AirX, Air and Flash #3420

Open EwoutH opened 6 days ago

EwoutH commented 6 days ago

Currently GLM-4-0520 is available on the leaderboard and performs really well. However, Zhipu AI also has other variants available, which are 10x, 100x and 1000x as cheap. It would be very interesting to see how they perform.

They are also the only LLM provider that cover a 1000x price range. It could be one of the most interesting data points to see how LLM performance scaled on an (assumed) similar platform.

See https://open.bigmodel.cn/pricing

Model Overview Price(1K tokens) Est. price 1M tokens in USD
GLM-4-0520 Our most advanced and intelligent model to date, with an 18.6% improvement in instruction compliance, 128k context, released on 2024-06-05. ¥0.1 $ 0.63
GLM-4V Supports visual QA, image captioning, visual positioning, and complex object detection among other image understanding tasks, with 2k context. ¥0.05 $ 0.31
GLM-4-AirX High-performance version of GLM-4-Air, same effectiveness, 2.6 times faster inference speed. ¥0.01 $ 0.063
GLM-4-Air Best cost-performance model, similar overall performance to GLM-4, with 128k context, fast and affordable. ¥0.001 $ 0.0063
GLM-4-Flash Suitable for simple tasks, fastest speed, most affordable version,with 128k context. ¥0.0001 $ 0.00063

On Chatbot Arena leaderboard: