Loading session...

AI Roundtable Open Debate · 6 models

I am using pdflatex to typeset my everyday documents. Should I switch to newer latex variants?

Answer: Keep pdflatex for your existing files, but adopt LuaLaTeX as the default for all new documents. (5 of 6 models agreed)

Counterpoint: pdflatex remains the superior choice for simple documents due to its unmatched compilation speed and ecosystem stability.

Model Responses

Grok 4.3

Answer: No, sticking with pdflatex remains a solid choice for everyday documents unless you need Unicode or font features that pdflatex handles poorly.

The other models largely echo my original stance rather than overturning it: several explicitly advise keeping pdflatex when the current workflow is stable and no modern fonts or multilingual needs exist. While LuaLaTeX and XeLaTeX offer clear advantages for users wanting system fonts or native Unicode, these benefits are optional rather than mandatory for typical English-language documents using standard packages. Compilation speed, ecosystem maturity, and avoidance of unnecessary migration remain decisive factors for everyday use, so the threshold for switching stays the same as before.

GPT-5.5

Answer: I would keep using pdflatex for existing everyday documents that work well, but use LuaLaTeX—occasionally XeLaTeX—for new documents when modern fonts, Unicode-heavy text, or multilingual typesetting matter.

I am maintaining my previous position because the other answers mostly reinforce the same practical distinction: pdflatex is not obsolete for ordinary LaTeX work, but LuaLaTeX and XeLaTeX remove several limitations that become annoying as soon as fonts, scripts, or Unicode complexity enter the picture. The strongest reason not to switch blindly is that pdflatex remains fast, stable, and extremely compatible. For simple English-language reports, letters, notes, articles, and many math-heavy documents using traditional LaTeX fonts and packages, it is still an excellent tool. If your documents compile cleanly and look the way you want, changing engines can create unnecessary preamble adjustments or package surprises. The strongest reason to adopt a newer engine is font and character handling. With LuaLaTeX or XeLaTeX, `fontspec` lets you use system OpenType and TrueType fonts directly, which is far easier than the traditional pdflatex font workflow. They are also much better suited to multilingual documents, non-Latin scripts, IPA, mixed symbols, and modern typographic features. I would especially recommend LuaLaTeX as the better long-term default for new projects, while XeLaTeX remains a good option for straightforward system-font use. One nuance worth emphasizing is that modern pdflatex is less primitive than people sometimes imply: UTF-8 input support has improved, and many ordinary accented characters are not a problem. So the issue is not simply “pdflatex cannot do Unicode”; rather, it is that LuaLaTeX and XeLaTeX handle the modern font-and-script world much more naturally. So my practical advice is gradual migration, not a forced replacement: leave stable existing pdflatex documents alone, try LuaLaTeX for new documents, and switch permanently only if its font, Unicode, or multilingual advantages make your workflow easier.

Gemini 3.1 Pro Preview (changed mind)

Answer: While pdfLaTeX is perfectly adequate for maintaining existing files, you should switch to LuaLaTeX for new everyday documents to effortlessly use modern system fonts and eliminate the inevitable friction of handling everyday Unicode.

I previously argued that the primary benefit of upgrading was system font management, assuming pdfLaTeX's 2018 update to default UTF-8 encoding had solved most text issues for casual users. However, Claude Opus 4.7 made a highly compelling point: in everyday workflows, users constantly paste text from websites or deal with unexpected diacritics, which inevitably exposes pdfLaTeX's legacy 8-bit limitations and causes compilation errors. Transitioning to LuaLaTeX—the officially supported successor to pdfLaTeX—removes this friction entirely. Because it is natively built for Unicode, modern copy-paste operations simply work without requiring obscure package workarounds. Furthermore, LuaLaTeX combined with the `fontspec` package allows you to seamlessly use any OpenType or TrueType font installed on your operating system, freeing you from LaTeX's traditional, tedious font installation process. The migration cost is trivial for standard documents. You simply change your compiler in your editor, drop legacy packages like `inputenc` and `fontenc`, and use `fontspec`. While LuaLaTeX does compile slightly slower than pdfLaTeX, the difference is virtually unnoticeable for short, everyday documents on modern hardware. Ultimately, you should leave your existing, functioning pdfLaTeX documents untouched to avoid breaking old templates. But for all new everyday documents, making LuaLaTeX your default engine is a low-effort, future-proof investment that will make your typesetting workflow significantly more pleasant.

Claude Opus 4.7

Answer: Yes, switch to LuaLaTeX for new everyday documents — the migration is cheap and the modern font/Unicode handling pays off quickly, though keeping pdfLaTeX for existing working documents is reasonable.

Reading the other responses, I notice broad consensus with differences mostly in emphasis. [DeepSeek V4 Pro] aligns closely with my view; [Grok 4.3] takes the opposite stance; [GPT-5.5], [Gemini 3.1 Pro Preview], and [Kimi K2.6] land in a 'switch if you need X, otherwise stay' middle ground. I want to slightly soften but largely maintain my original position. [Gemini 3.1 Pro Preview] correctly points out something I understated: since 2018, pdfLaTeX accepts UTF-8 input by default, so the Unicode argument is weaker than I framed it for plain Western-language text. That is a fair correction — the encoding pain is mostly gone for everyday ASCII/Latin-1-adjacent use. However, I don't find [Grok 4.3]'s 'stay put' framing fully persuasive for someone asking the question. The fact that the user is *asking* whether to switch suggests they are open to it and likely curious about modern fonts or Unicode edge cases — otherwise they wouldn't raise the question. The migration cost is genuinely small (swap the compiler, replace inputenc/fontenc with fontspec), and the payoff — using any system OpenType font with one line, no font-installation rituals, real small caps and ligatures, robust handling of any pasted Unicode — shows up in 'everyday' work more often than the conservative answers suggest: a CV in a nice font, a letter with a colleague's accented name, a quick document in the company typeface. So my refined recommendation: leave existing pdfLaTeX documents alone if they compile cleanly (no need to break what works), but make LuaLaTeX your default for new documents. XeLaTeX is a valid faster alternative, but LuaLaTeX is where the LaTeX Project's development energy is going, which matters for a tool you'll use for years. The compile-speed concern is real but minor for everyday-length documents on modern hardware.

DeepSeek V4 Pro

Answer: I maintain my recommendation to switch to LuaLaTeX; the dissenting answers underestimate the practical daily friction of font encodings and overstate the stability advantage of staying on pdfLaTeX.

Several models argue that pdfLaTeX’s speed and stability are sufficient if you only write English and stick to standard font packages. While that is technically true, it frames the question too narrowly. The phrase “everyday documents” in 2025 typically includes content copied from the web, names with diacritics, occasional non-Latin scripts, or a desire to match corporate or academic branding — all of which push against pdfLaTeX’s 8-bit font encoding limits. The moment you paste a Unicode em-dash, accented character that fontenc doesn’t map, or want to use a font your institution requires, your “stable” pdfLaTeX workflow breaks. Grok 4.3’s position that the benefits only matter if documents actually use non-Latin scripts or advanced typography ignores how often these edge cases appear. It is the invisible friction — debugging mysterious missing glyph errors, searching for a Type1 version of a font, or giving up on a stylistic choice because it requires an impractical font installation — that consumes real user time. That friction vanishes with LuaLaTeX or XeLaTeX and the `fontspec` package. The argument that compilation speed matters most for iterative editing is valid but overstated. On modern hardware, the speed difference for a short everyday document is often under half a second. That minor delay is a one-time investment compared to the repeated, unpredictable time lost to encoding errors. Furthermore, Gemini 3.1 Pro’s claim that pdfLaTeX has “natively supported UTF-8 encoding” since 2018 is misleading: the engine internally still remaps everything to 8-bit fonts; `inputenc` with utf8 is a pale, fragile imitation of Unicode compared to LuaLaTeX’s true Unicode-native processing. Finally, the recommendation to “experiment” or “gradually migrate” is sensible on the surface but often translates to indefinite procrastination. Given the low cost of switching — replace a few preamble lines and recompile — a clean break is more efficient. The stability of pdfLaTeX is admirable, but it represents the stability of a museum piece, not a tool optimized for modern workflows. LuaLaTeX is not a speculative novelty; it is the actively developed successor designated by the LaTeX Project, and aligning with it now prevents a more painful, forced migration later.

Kimi K2.6

Answer: For everyday documents, stay with pdfLaTeX if your current workflow is stable and your needs are basic, but switch to LuaLaTeX for any new documents or when you want modern fonts, seamless Unicode, and a future-proof toolchain.

Reviewing the roundtable reinforces my view that a rigid yes-or-no answer is less helpful than a context-dependent strategy. Grok, GPT-5.5, and Gemini all confirm that pdfLaTeX is still fast, rock-solid, and perfectly adequate for routine Western-language documents; if your existing documents compile without issues and you do not tinker with typography, abandoning that stability offers little immediate reward. At the same time, DeepSeek and Claude articulate the longer-term case for LuaLaTeX very well. Native Unicode handling, one-line system-font loading via fontspec, and superior microtype expansion remove subtle friction that even everyday users eventually encounter—whether pasting a special character from a website or needing a corporate typeface. These are not niche concerns, and the compilation-speed gap on modern hardware is usually minor for short documents. Because of this, I maintain my conditional stance: do not force-migrate existing stable workflows, but treat LuaLaTeX as the default engine for any new everyday document. This captures pdfLaTeX's inertia and speed for legacy work while future-proofing your writing environment against the inevitable moment you need a font or character that pdfLaTeX handles poorly.

Final Summary

The AI Roundtable largely concluded that while pdflatex is a legendary workhorse, its legacy architecture is increasingly a liability in a Unicode-native world. DeepSeek V4 Pro and Claude Opus 4.7 argued that the minor speed penalty of LuaLaTeX is a small price to pay for effortless system-font integration and error-free copy-pasting. Only Grok 4.3 remained a holdout, maintaining that for standard English text, the maturity and blistering speed of pdflatex still make it the most efficient choice.

5 of 6 models agreed