New Free AI Tess-4.5 Thinks Less and Answers Faster

A new AI model called Tess-4.5-27B has been released for free on Hugging Face. It's a fine-tuned version of the Qwen model, created by the developer behind Synthia. The key improvement is efficiency: it reaches the same performance as its predecessor but uses 25% fewer tokens—the units of text that AI processes. Fewer tokens mean faster answers, lower computing costs, and the ability to handle longer conversations before running out of memory.

The efficiency gains are even bigger on tough tasks. In coding challenges, it used 32% fewer tokens and still scored higher. For data extraction, it used 43% fewer tokens. Overall, across 150 test scenarios, it generated 262,000 tokens versus 349,000 for the base model. This is like a car that goes the same distance on a quarter less gas.

For everyday users, this means AI assistants can respond more quickly and cheaply. If you use AI for work or personal projects, you'll spend less on API calls or be able to do more within free tiers. On your own computer, the savings compound: fewer tokens and each token processed faster, so you get answers sooner and can have longer sessions.

While this is a technical release aimed at developers, the benefits trickle down to anyone who uses AI tools. More efficient models make AI more accessible and affordable. The model is available for free, so developers can experiment and build new applications. As AI becomes more efficient, we can expect faster, cheaper, and more capable assistants in the apps we use daily.