Tencent releases Hy3 open-source model that allegedly matches models up to five times its active size
Tencent has released an open-source language model called Hy3, which is surprisingly efficient despite its large size. Hy3 has 295 billion parameters, but only 21 billion are active at a time.
What happened
Hy3 uses a unique architecture that allows it to match models two to five times its size, while reducing the rate of 'hallucinations' (errors) by half. This is impressive because large language models often struggle with efficiency and accuracy. The model's developers claim that only 5.4% of its responses are incorrect.
Why it matters
For businesses that rely on language models, Hy3's efficiency and accuracy could be a game-changer. This means you might be able to use a smaller, more manageable model to achieve the same results as a much larger one. You could also save on computational resources and costs.
The takeaway
You may want to consider exploring open-source language models like Hy3, which could offer a more efficient and cost-effective solution for your business needs.
Our plain-English take, written from public reporting for operational business owners. Always read the original for full context.
Nayre builds the AI systems behind stories like this.
Chatbots, workflow automation, finance intelligence, and internal knowledge systems. Built for operational teams, shipped in days.