welm

#welm

@Fenng: This paper from WeChat's WeLM team reveals at least two model scales: 80B and 617B. Each scale includes a standard version and an HD4 version, namely WeLM-80B, WeLM-HD4-80B, WeLM-617B, and WeLM-HD4-617B. Among them, WeLM-617…

X AI KOLs Timeline ↗ · 2026-07-12 Cached

The WeChat WeLM team published a paper introducing the Hidden Decoding method, which extends computation through hidden flows without increasing the Transformer backbone parameters, training the WeLM-HD4-80B and WeLM-HD4-617B MoE models, surpassing autoregressive baselines on multiple benchmarks.

0 favorites 0 likes

#welm

@Fenng: Saw this piece written by a self-media account — 'The latest fourth-generation WeLM-80B now has only 80 billion total parameters, with 3 billion activated, an activation rate of just 3.75%. For comparison — DeepSeek-V4-Flash, the domestic representative of extreme cost-performance, has 284 billion total parameters, 13 billion activated, activation rate of 4.6%...'

X AI KOLs Timeline ↗ · 2026-06-23 Cached

Fenng shares a self-media comparison between the fourth-generation WeLM-80B (80B total params, 3B activated, 3.75% activation rate) and DeepSeek-V4-Flash (284B total, 13B activated, 4.6% activation rate), with a humorous comment.

0 favorites 0 likes

#welm

@Fenng: WeLM is a proprietary large model developed by Tencent's WeChat team, while Hunyuan is Tencent's open-source large model; they have different positioning. Since it's a closed-source large model, it cannot disclose technical details externally, nor can it participate in evaluations (evaluations are just a promotional tool). If you want to pay attention, it's better to experience it through the product.

X AI KOLs Following ↗ · 2026-06-22

Fenng explained the positioning difference between Tencent's WeLM (closed-source large model) and Hunyuan (open-source large model), pointing out that the closed-source model will not disclose technical details or participate in evaluations, and suggested understanding it through product experience.

0 favorites 0 likes

#welm

@Fenng: The WeLM large model developed by the WeChat team has likely entered the top tier of domestically produced large models.

X AI KOLs Following ↗ · 2026-06-22

The WeLM large model developed by the WeChat team is considered to have entered the top tier of domestically produced large models.

0 favorites 0 likes

#welm

@Fenng: WeChat's AI Agent has arrived, product name 'Xiao Wei', main model uses WeLM, partial answers fallback to DeepSeek, grayscale testing has begun.

X AI KOLs Following ↗ · 2026-06-20

WeChat launches AI Agent product 'Xiao Wei', main model uses WeLM, some answers fallback to DeepSeek, grayscale testing has begun.

0 favorites 0 likes

welm

@Fenng: This paper from WeChat's WeLM team reveals at least two model scales: 80B and 617B. Each scale includes a standard version and an HD4 version, namely WeLM-80B, WeLM-HD4-80B, WeLM-617B, and WeLM-HD4-617B. Among them, WeLM-617…

@Fenng: The WeLM large model developed by the WeChat team has likely entered the top tier of domestically produced large models.

@Fenng: WeChat's AI Agent has arrived, product name 'Xiao Wei', main model uses WeLM, partial answers fallback to DeepSeek, grayscale testing has begun.

Submit Feedback