# LMArena 39 Hacker News commenters recommended LMArena and 40 criticized it, November 2025 to November 2026. It was recommended in 11 of 13 months. **Verdict:** People recommend it as a leaderboard built on blind human head-to-head votes that is harder to game than benchmarks. Critics say the votes reward style and sycophancy over factual accuracy. ## Why people recommend it - Blind human head-to-head comparisons (5 comments): "On paper it's one of the best because it's meant to be blind comparison of your own prompts." ([source](https://news.ycombinator.com/item?id=48313569)) "they use humans to judge a head-to-head comparison between a couple models at a time" ([source](https://news.ycombinator.com/item?id=49336521)) "it helps to find a anonymously a model without confirmation bias" ([source](https://news.ycombinator.com/item?id=48581510)) - Useful overview of many models (4 comments): "having all these frameworks compared in one place is actually pretty useful" ([source](https://news.ycombinator.com/item?id=47650520)) "every comparison like this or at models.dev or arena.ai is good" ([source](https://news.ycombinator.com/item?id=47695792)) "I've found this company is a pretty good ranker" ([source](https://news.ycombinator.com/item?id=48312054)) - Harder to game than benchmarks (3 comments): "IMO LMArena is the best benchmark that avoids benchmaxxing" ([source](https://news.ycombinator.com/item?id=48849243)) "Lmarena is a bit harder to fake and is human-evaluated." ([source](https://news.ycombinator.com/item?id=47617034)) "it's the best benchmark that avoids benchmaxxing" ([source](https://news.ycombinator.com/item?id=48865639)) ## What people warn about - Rewards style and tone over accuracy (4 comments): "Main issue with it these days are that it kinda measures sycophancy and user preferred tone more than substance." ([source](https://news.ycombinator.com/item?id=46240780)) "it has been indicative of which LLMs style I like best, not necessarily its factual accuracy" ([source](https://news.ycombinator.com/item?id=46515167)) "Models that train on Arena converge on helpfulness, not necessarily truthiness" ([source](https://news.ycombinator.com/item?id=48132143)) - Shallow side-by-side votes miss complex tasks (4 comments): "It's biased to small context performance, which is why I don't pay much attention to it as a developer aside from a quick glance." ([source](https://news.ycombinator.com/item?id=45857932)) "biased towards visual aesthetics more than raw coding as the voters are just asked which generated website they prefer" ([source](https://news.ycombinator.com/item?id=48316137)) "the methodology of letting random people compare outputs side-by-side is a very shallow judgement method in my opinion" ([source](https://news.ycombinator.com/item?id=48990917)) - Prompt mix unlike real-world use (2 comments): "the distribution of prompts on arena.ai can be pretty different than real-world usage" ([source](https://news.ycombinator.com/item?id=50262416)) "it only captures chat, and relies on users spending their time improving it" ([source](https://news.ycombinator.com/item?id=50214353)) - Gaming and benchmaxxing concerns (3 comments): "It's easy to game and human evaluation data has its trade-offs" ([source](https://news.ycombinator.com/item?id=47617896)) "Arena can definitely be benchmaxxed a bit, if you try." ([source](https://news.ycombinator.com/item?id=48866047)) "There's a ton of arenamaxxing going on, though I don't disagree that it's one of the better actual benchmarks." ([source](https://news.ycombinator.com/item?id=48865908)) ## Alternatives hackers recommend - [Artificial Analysis](https://hackerapproved.com/p/artificial-analysis.md): 62 recommend - [ARC-AGI](https://hackerapproved.com/p/arc-agi-3.md): 22 recommend - [OCR Arena](https://hackerapproved.com/p/ocr-arena.md): 11 recommend - [Bullshit Benchmark](https://hackerapproved.com/p/bullshit-benchmark.md): 9 recommend - [GenAI Showdown](https://hackerapproved.com/p/genai-showdown.md): 9 recommend ## Where to get it - [Visit the official site](https://arena.ai/) ## Cite as Hacker Approved, "LMArena", https://hackerapproved.com/p/lmarena/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).