Arena · XUpdated Original · English

Jev prioritised efficiency and routed most calls to DeepSeek V4.1 Flash - over a quarter (28.1%) of all model calls.

Image source · Arena · X

Jev prioritised efficiency and routed most calls to DeepSeek V4.1 Flash - over a quarter (28.1%) of all model calls.

It showed preference for OpenAI models over Anthropic’s LLMs. GPT-6 Astra was the second most popular model (23.5%), followed by other GPT models, Sol (22.8%) and Luna (11.1%).

Original source

Arena · X

Content notes

Original publication and rights belong to the source.