Introducing Hy4 Preview

<p><strong><a href="https://hy.tencent.ai/research/hy4-preview">Introducing Hy4 Preview</a></strong></p> New open weight text input (no vision) LLM from Chinese company Tencent today: 770B total parameters, 49B active parameters, 1M token context window, <a href="https://huggingface.co/tencent/Hy4-preview">1.56TB on Hugging Face</a>.</p> <p>This is a big size increase from their previous <a href="https://huggingface.co/tencent/Hy3">Hy3</a> in July, which was 295B, 21B active, 256,000 context, 598GB.</p> <p>I recently started using model chat templates to better understand their capabilities. Here's Hy4's <a href="https://huggingface.co/tencent/Hy4-preview/blob/main/chat_template.jinja">chat_template.jinja</a> on Hugging Face, which includes this section:</p> <div class="highlight highlight-text-html-django"><pre><span class="pl-e">{%</span>- <span class="pl-k">if</span> <span class="pl-k">not</span> <span class="pl-s">reasoning_effort</span> <span class="pl-s">is</span> <span class="pl-s">defined</span> <span class="pl-e">%}</span> <span class="pl-e">{%</span>- <span class="pl-s">set</span> <span class="pl-s">reasoning_effort</span> = <span class="pl-s">'high'</span> <span class="pl-e">%}</span> <span class="pl-e">{%</span>- <span class="pl-s">elif</span> <span class="pl-s">reasoning_effort</span> <span class="pl-k">not</span> <span class="pl-k">in</span> [<span class="pl-s">'high'</span>, <span class="pl-s">'no_think'</span>] <span class="pl-e">%}</span> <span class="pl-e">{%</span>- <span class="pl-k">if</span> <span class="pl-s">reasoning_effort</span> <span class="pl-s">is</span> <span class="pl-s">none</span> <span class="pl-e">%}</span> {{- raise_exception('reasoning_effort error : None, should be no_think/high') }} <span class="pl-e">{%</span>- <span class="pl-k">else</span> <span class="pl-e">%}</span> {{- raise_exception('reasoning_effort error : ' + reasoning_effort + ', should be no_think/high') }} <span class="pl-e">{%</span>- <span class="pl-k">endif</span> <span class="pl-e">%}</span> <span class="pl-e">{%</span>- <span class="pl-k">endif</span> <span class="pl-e">%}</span></pre></div> <p>So it looks like there are just two reasoning effort levels: "high" (the default) and "no_think" (reason by disabled).</p> <p>I tried my "Generate an SVG of a pelican riding a bicycle" prompt with the default high reasoning <a href="https://openrouter.ai/tencent/hy4-preview#apps">via OpenRouter</a> and <a href="https://tools.simonwillison.net/markdown-svg-renderer#url=https%3A%2F%2Fgist.github.com%2Fsimonw%2Fcb69816b3fb940f2782569a82a523af1">got this</a>:</p> <p><img alt="Flat vector cartoon illustration of a white pelican with a large orange bill riding a red bicycle to the right along a grey road with a dashed white centre line, its orange webbed feet on the pedals and grey tail feathers fanned out behind, against a pale blue sky with a yellow sun, white clouds and horizontal white motion lines suggesting speed" src="https://static.simonwillison.net/static/2026-08-29/IMG_7725.jpeg" /></p> <p>Quoting the reasoning trace:</p> <blockquote> <p>[...] 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.</p> <p>Maybe add sunglasses? no.</p> <p>Maybe add water? no.</p> </blockquote> <p>It's interesting how the reasoning trace uses slightly truncated English, presumably because perfect grammar isn't useful or token efficient for hidden reasoning text. <p>Tags: <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/pelican-riding-a-bicycle">pelican-riding-a-bicycle</a>, <a href="https://simonwillison.net/tags/llm-reasoning">llm-reasoning</a>, <a href="https://simonwillison.net/tags/llm-release">llm-release</a>, <a href="https://simonwillison.net/tags/ai-in-china">ai-in-china</a></p>