Inkling-Small AutoEval Scores Released on Text Arena
The evaluation platform Text Arena has released the AutoEval scores for Inkling-Small, a new open-weights multimodal model developed by Thinking Machines Lab. This evaluation allows developers to compare Inkling-Small's performance against other LLMs in real-world benchmarks, highlighting the capabilities of smaller, efficient Mixture-of-Experts (MoE) models. Inkling-Small is a multimodal MoE model that accepts text, image, and audio inputs, offering controllable reasoning effort at a quarter of the size of the original Inkling model.
## BACKGROUND
Text Arena is a platform and leaderboard for evaluating and comparing large language models across tasks like coding, math, and creative writing. Inkling-Small is an open-weights model designed by Thinking Machines Lab to deliver high performance and multimodal capabilities with reduced computational requirements.