腾讯发布 Hy4 预览版：770B 参数开源文本模型
Introducing Hy4 Preview
Introducing Hy4 Preview
介绍Hy4预览版
New open weight text input (no vision) LLM from Chinese company Tencent today: 770B total parameters, 49B active parameters, 1M token context window.
今天，中国公司腾讯发布了一款新的开放权重文本输入（无视觉）大语言模型：总参数770B，激活参数49B，上下文窗口1M token。
This is a big size increase from their previous Hy3 in July, which was 295B, 21B active, 256,000 context.
与7月份发布的Hy3相比，这是一个大幅度的规模提升，Hy3总参数295B，激活参数21B，上下文窗口256,000。
I recently started using model chat templates to better understand their capabilities. Here's Hy4's chat_template.jinja on Hugging Face, which includes this section:
我最近开始使用模型聊天模板来更好地理解它们的能力。这是Hugging Face上Hy4的chat_template.jinja，其中包括以下部分：
{%- if not reasoning_effort is defined %}
{%- set reasoning_effort = 'high' %}
{%- elif reasoning_effort not in ['high', 'no_think'] %}
{%- if reasoning_effort is none %}
{{- raise_exception('reasoning_effort error : None, should be no_think/high') }}
{%- else %}
{{- raise_exception('reasoning_effort error : ' + reasoning_effort + ', should be no_think/high') }}
{%- endif %}
{%- endif %}{%- if not reasoning_effort is defined %}
{%- set reasoning_effort = 'high' %}
{%- elif reasoning_effort not in ['high', 'no_think'] %}
{%- if reasoning_effort is none %}
{{- raise_exception('reasoning_effort error : None, should be no_think/high') }}
{%- else %}
{{- raise_exception('reasoning_effort error : ' + reasoning_effort + ', should be no_think/high') }}
{%- endif %}
{%- endif %}So it looks like there are just two reasoning effort levels: "high" (the default) and "no_think" (reason by disabled).
所以看起来只有两个推理努力级别："high"（默认）和"no_think"（禁用推理）。
I tried my "Generate an SVG of a pelican riding a bicycle" prompt with the default high reasoning via OpenRouter and got this:
我尝试了通过OpenRouter使用默认高推理的"生成一个骑自行车的鹈鹕的SVG"提示，得到了这个结果：
Quoting the reasoning trace:
引用推理轨迹：
[...] Let's maybe add a helmet? It could improve riding theme, but may obscure head. Maybe a small cycling cap or helmet? The user didn't ask; can add red helmet? Might be cute. But pelican with big beak; a helmet might obscure. Better maybe no.
[...] 也许加个头盔？它可以增强骑行主题，但可能会遮挡头部。也许加个小骑行帽或头盔？用户没有要求；可以加个红色头盔？可能很可爱。但鹈鹕有巨大的喙；头盔可能会遮挡。也许最好不要。
Maybe add sunglasses? no.
也许加个太阳镜？不。
Maybe add water? no.
也许加水？不。
It's interesting how the reasoning trace uses slightly broken English, presumably because perfect grammar isn't useful or token efficient for hidden reasoning text.
有趣的是，推理轨迹使用了略显蹩脚的英语，大概是因为完美的语法对隐藏推理文本来说并不实用或节省token。
Tags: ai, generative-ai, llms, pelican-riding-a-bicycle, llm-reasoning, llm-release, ai-in-china
标签：ai, generative-ai, llms, pelican-riding-a-bicycle, llm-reasoning, llm-release, ai-in-china
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力