Track HN: Comparing 156 LLM Launch Posts on Hacker News
A new analytics report from Track HN has analyzed 156 LLM launch posts on Hacker News spanning March 2023 to April 2026, revealing which AI providers dominate community attention and how engagement patterns differ wildly across companies. The findings show Anthropic punches far a...
According to Track HN, the analytics platform behind the report published April 13, 2026 and updated April 15, 2026, the data covers 132 distinct model versions across 24 model families from 13 providers. The full dataset captures 56,041 total comments across those posts, with 735 non-deleted top-level comments manually labeled for sentiment. This is one of the more rigorous third-party attempts to quantify how the Hacker News technical community actually responds to AI model releases, not just how many people clicked upvote.
Why This Matters
Hacker News is not Twitter. The people commenting there are engineers, researchers, and founders who actually build things with these models, and their collective reaction to a launch shapes early adoption in ways that press releases never could. Anthropic achieving the highest average score per post at 1,254, across just 11 posts, tells you something real about how the technical community ranks trust and quality. The fact that xAI sits at a score-to-comment ratio of 0.9, the only provider below 1.0, means Elon Musk's AI lab generates controversy faster than appreciation, which is a credibility problem that matters for enterprise adoption.
Daily briefing from 50+ sources. Free, 5-minute read.
The Full Story
Track HN built this report by pulling every Hacker News post tied to a model launch or release that crossed a minimum threshold of 50 points and 50 comments. Posts had to point to official announcements, GitHub repositories, research papers, or credible third-party coverage. Duplicates got collapsed, and the whole thing was verified by hand. The result is a dataset covering posts from March 2023 through April 2026, capturing the period when LLM releases went from rare events to routine calendar items.
The provider rankings tell a story about volume versus quality. OpenAI leads with 28 posts and a total score of 22,602, but its score-to-comment ratio sits at 1.35, meaning it gets decent engagement but not exceptional discussion quality relative to its score. Google DeepMind posted 24 times and accumulated 21,055 total points at a ratio of 1.91. Both companies dominate by raw numbers because they release models constantly. Anthropic, by contrast, posted just 11 times but averaged 1,254 points per post, the highest of any provider with more than three posts, and a score-to-comment ratio of 1.94.
Meta is the sleeper in this dataset. Six posts, 7,488 total points, and an average score of 1,248 per post, nearly matching Anthropic's average with fewer than half the launches. That reflects the genuine excitement in the open-source community every time a new Llama variant drops. DeepSeek, with 10 posts and a 2.28 ratio, also outperforms its post count, which aligns with the broader narrative around its cost-efficiency breakthroughs that dominated AI conversation in early 2025.
The concentration data is worth sitting with. The top 5 posts account for 13.5 percent of all score across all 156 posts. That means a small number of landmark releases, likely GPT-4, Claude 3, Llama 3, or DeepSeek R1 territory, captured a disproportionate share of attention. The rest of the 156 posts fought over the remaining 86.5 percent. This mirrors how media attention works in every industry, but it confirms that most model launches are noise to the Hacker News crowd.
The methodology for sentiment labeling used an LLM to classify the top 5 comments by display order from each post, producing 735 labeled samples total. Themes were identified via keyword matching. This is a reasonable approach for a dataset this size, though it means the sentiment data reflects early comment reactions, which on Hacker News tend to be from the most technically engaged readers who arrive first. Whether that represents the full range of community opinion is debatable, but it is probably more meaningful than upvote counts alone.
September 2025 was the single busiest month in the dataset, with 11 distinct model versions launching that month alone. That peak suggests either a coordinated release cycle among providers or a natural inflection point in the competitive race to ship before end-of-year deadlines.
Key Details
- The dataset covers 156 posts from 13 providers, spanning March 2023 to April 2026.
- Total comments across all posts reached 56,041, with 735 top-level comments labeled for sentiment.
- Anthropic averaged 1,254 points per post, the highest average among providers with more than 3 posts.
- OpenAI posted 28 times for 22,602 total points, but its score-to-comment ratio of 1.35 is below average for the group.
- xAI posted 7 times and earned a score-to-comment ratio of 0.90, the only provider below 1.0.
- September 2025 saw 11 model versions launch in a single month, the peak in the dataset.
- The top 5 posts across all 156 account for 13.5 percent of total score.
- Microsoft posted just 5 times but achieved a 3.46 score-to-comment ratio, suggesting its posts spark little debate but attract approving readers.
What's Next
Track HN has flagged several features still in development, including topic detection for emerging themes and full sentiment trend analysis across a post's full comment lifespan rather than just the top 5 early comments. As the platform matures, expect the sentiment data to become more granular, potentially revealing whether specific model capabilities like coding or reasoning generate more positive reactions than general-purpose benchmarks. Providers with low engagement efficiency ratios, particularly xAI and Cohere, should watch whether their next launches shift community perception or continue the same patterns.
How This Compares
This report sits alongside a small but growing category of Hacker News analytics tools, including Algolia's full-text HN search index and various personal projects that track specific keywords or users. What separates Track HN's LLM report from those efforts is domain specificity. It is not tracking all of Hacker News. It is tracking one specific class of announcement and building a longitudinal record of how community reaction has evolved over three years. That focus makes the data more actionable for anyone building or launching AI tools and platforms.
Compare this to the broader conversation happening on Hacker News itself. A separate post asking whether LLM benchmark improvements are actually plateauing earned 174 points and 156 comments, suggesting the community is increasingly skeptical of launch-day claims. Track HN's data captures that skepticism quantitatively: providers with high comment-to-score ratios tend to generate more debate than celebration. The xAI result is the clearest example, but even OpenAI's relatively modest 1.35 ratio hints at a more contested reception than its marketing would suggest.
The timing of this report also matters. At 11 model launches in a single month during September 2025, the pace of releases has clearly reached a saturation point. For anyone following AI news day to day, this dataset is useful confirmation that community attention is a finite resource, and most launches are competing for scraps while a handful of genuinely surprising releases capture everything.
FAQ
Q: What is Hacker News and why do AI companies care about it? A: Hacker News is a technology discussion forum run by Y Combinator. It attracts software engineers, researchers, and startup founders who discuss and critique new technology. AI companies pay attention to it because positive reception there often translates to early adoption among the technical practitioners who build products using these models.
Q: How did Track HN decide which posts to include in the 156? A: Posts had to clear a minimum threshold of 50 points and 50 comments, be tied to an actual model launch or release, and come from official sources, GitHub repositories, academic papers, or credible third-party coverage. The list was also verified by manual review to exclude duplicates.
Q: Why does Anthropic score so high per post despite fewer launches than OpenAI? A: Anthropic releases models less frequently, which means each release carries more novelty and community interest. Its Claude models have also built a strong reputation among developers for reliability and instruction-following, so announcements tend to land with an audience already primed to engage positively.
The Track HN LLM report is a useful benchmark for understanding which AI providers are actually winning the technical community's trust, separate from mainstream press coverage or social media hype. As model launches continue at pace and community attention becomes harder to earn, data like this will matter more for teams thinking strategically about how and when to ship. Subscribe to the AI Agents Daily weekly newsletter for daily updates on AI agents, tools, and automation.
Get stories like this daily
Free briefing. Curated from 50+ sources. 5-minute read every morning.
