hn.today

Gemini 4 Neon

blog.google1030 points765 comments
Screenshot of Gemini 4 Neon

Commenters see Neon as another step in Google’s return to frontier-model competition, but disagree about what its strengths mean. Some cite benchmark results and unusually high speed as evidence that it offers near-top quality with better latency, and expect it to become a useful default. Others see the model lineup as specialized: Neon may score better overall, while Carbon remains preferable for long, research-heavy reasoning or complex coding. A Carbon user describes major gains on a difficult engineering project, while another says Opus 5.5 feels more like a capable coworker and stays oriented across extended tasks in a way Gemini models do not. These comparisons highlight a divide between benchmark performance and sustained, practical usefulness; some also praise Gemini’s writing style and Google’s rapid releases.

Many commenters welcome Google’s momentum, yet find its product choices hard to navigate. The element-based names, overlapping model tiers, unclear guidance on which model to use, and inconsistent subscription or Antigravity access frustrate users. Some are unsure whether Neon replaces Carbon or simply sits between faster and deeper-reasoning options; others object to legacy models disappearing without warning, especially when established workflows depend on them. Cost and speed also shape evaluations: Neon’s performance may justify its price for some, while others want top-tier capability at lower cost. One commenter says they will avoid Gemini until Google’s AI-training opt-out reaches their account, citing concern about connected personal data.

Read on blog.google765 comments on Hacker News

Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.

More in AI

Mapping every neuron in Qwen3-8B

Mapping every neuron in Qwen3-8B

Researchers found individual neurons in Qwen3-8B that reliably steer its replies into distinct personas or styles, including pirate, German-speaking, and childlike voices. Nudging these neurons changed outputs across many questions without prompting or fine-tuning, revealing how specific behaviors can be localized and edited inside a language model. (fermionresearch.com)

ChatGPT for SNES

ChatGPT for SNES

ChatGPT for SNES runs a 2.2-million-parameter language model on the Super Nintendo’s Super FX 2 chip, with an optional Wikipedia fact book. The unofficial experiment can run in an emulator or on a flash-cart-equipped console, but its tiny model often gives incorrect answers. (bxfoundry.com)

The daily digest

Today's best Hacker News stories, summarized and screenshotted, one email a day.