Tencent Mixyuan Releases Hy3 1bit and 4bit Compressed Versions

Tencent Mixyuan has released the 1bit and 4bit compressed versions of Hy3, significantly lowering the deployment threshold for the 295 billion parameter model. The original weight of Hy3 is nearly 600GB, while the 1bit version has been reduced to 85.5GiB, allowing it to run on a single 96GB NVIDIA H20 graphics card. When running on a single card, users need to disable some acceleration features and shorten the text length processed by the model. After using two H20 graphics cards, acceleration can be enabled. The 4bit version has a size of 169.9GiB and requires at least two H20 graphics cards; although it occupies more video memory, its performance is closer to the original model. Tencent Mixyuan stated that the 4bit version is comparable to the original model in terms of code, tool invocation, long text understanding, and Agent tasks, with the 1bit version showing only a slight decrease. With acceleration enabled, the 1bit version's speed increases by about 50%, while the 4bit version sees an improvement of nearly 60%. Additionally, Tencent Mixyuan has also provided a 4bit version suitable for server deployment, which can be made into an API for multiple users or applications to call.

Disclaimer: This content is provided for general branding and informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online events, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets or to use any services. Crypto assets are highly volatile and may result in loss. WEEX services and online events may not be available in all regions and are subject to applicable laws, regulations, and eligibility requirements. You are responsible for ensuring that your use of WEEX services complies with local laws and for carefully assessing the risks before participating in any crypto-related activities.