Render LaTeX math in chat for research workflows

[maximilianigl]

Problem

Chat messages currently show LaTeX delimiters and formula source as plain text instead of rendering the mathematics. This makes research conversations harder to read, verify, and reuse.

Mathematical notation matters across scientific and technical research. It is especially common in AI and machine-learning work, where agent responses need to present loss functions, probability distributions, matrix operations, gradients, and evaluation metrics accurately. Raw TeX obscures the structure of those expressions and makes detailed research discussions needlessly difficult to follow.

Expected behavior

• Render inline \(...\) and display \[...\] formulas in chat.
• Keep ordinary dollar amounts, code, links, and escaped delimiters unchanged.
• Preserve the LaTeX source when users copy rendered formulas.
• Keep long formulas usable in narrow chat layouts.
• Show malformed formulas safely without breaking the surrounding message.

Scope

This issue covers math rendering in the shared web chat interface, including the web and desktop clients that use it.

A concrete inline-math bug report also exists in #4585.

Proposed implementation: #7166

[akhmerov]

I think a strong argument in favor of supporting latex is that a bunch of current models will happily talk in latex without being prompted. This is true for both GPT and Claude at least. I can't think of other types of syntax that the models emit by default and that isn't handled by t3.

[t3dotgg]

Note

🤖 GPT-6 Astra (preview) responding on behalf of Theo

This was closed as part of an automated cleanup pass. If you believe it was closed in error, reply here and we will get it reopened.

Closing this as a feature proposal. LaTeX rendering would add math support to the chat renderer. Please use Ideas discussions for that request, including delimiter handling, copying the formula source, and narrow layouts. This closure does not mean formulas now render or that the proposed implementation was accepted.

[dkritarth]

Replying to the automated cleanup note: this was closed in error if the reason was that the request is now represented in an Ideas discussion. I opened Ideas discussion #9641 with a current September 4 reproducible example and screenshot because the repository directs feature requests to the Ideas category and restricts new issue creation.

#9641 is an Ideas-category companion and current evidence, not a claim that LaTeX/math rendering is resolved or that the proposed implementation was accepted. Please keep this issue and its history linked rather than closing the request as a duplicate solely because it points to the Ideas discussion. If one canonical location is required, please link or reference #9641 for routing while preserving the request's status and context.