PulseAugur
LIVE 16:25:54
significant · [5 sources] · · 日本語(JA) Gemini 3.5 Flashの総パラメータ数は2500億~3000億か、Hacker NewsユーザーがTPU性能から逆算 https:// fed.brid.gy/r/https://gigazine .net/news/20260520-gemini-3-5-flash-parameter/
4
significant

Gemini 3.5 Flash: High Cost, Fast Performance, and User Estimates

Google's new Gemini 3.5 Flash model is being discussed on Hacker News, with users estimating its total parameters at 250-300 billion and active parameters at 10-16 billion. Despite its "Flash" name, the model is reportedly significantly more expensive than previous versions, with one user noting a single SVG generation cost approximately $0.13. While praised for speed and performance in coding and single-shot reasoning tasks, Gemini 3.5 Flash appears to struggle with structural errors and long-term agent tasks, often adding decorative elements instead of fixing fundamental issues. AI

Summary written by gemini-2.5-flash-lite from 5 sources. How we write summaries →

IMPACT Gemini 3.5 Flash offers high speed and performance but at a significantly increased cost, prompting users to carefully consider its application for specific tasks.

RANK_REASON This cluster discusses a new model release from a major AI lab (Google) and includes user-driven analysis and discussion of its performance and cost.

Read on Mastodon — mastodon.social →

Gemini 3.5 Flash: High Cost, Fast Performance, and User Estimates

COVERAGE [5]

  1. Simon Willison TIER_1 Türkçe(TR) ·

    llm-gemini 0.32

    <p><strong>Release:</strong> <a href="https://github.com/simonw/llm-gemini/releases/tag/0.32">llm-gemini 0.32</a></p> <blockquote> <ul> <li>New model <code>gemini-3.5-flash</code> for <a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5/">Ge…

  2. Simon Willison TIER_1 Italiano(IT) ·

    llm-gemini 0.32a0

    <p><strong>Release:</strong> <a href="https://github.com/simonw/llm-gemini/releases/tag/0.32a0">llm-gemini 0.32a0</a></p> <blockquote> <ul> <li>Compatible with <code>llm&gt;=0.32a0</code> alpha - adds the ability to stream reasoning tokens.</li> </ul> </blockquote> <p>Tags: <a hr…

  3. Mastodon — mastodon.social TIER_1 日本語(JA) · [email protected] ·

    Gemini 3.5 Flash's total parameter count is 250 billion to 300 billion, Hacker News users calculate backwards from TPU performance

    Gemini 3.5 Flashの総パラメータ数は2500億~3000億か、Hacker NewsユーザーがTPU性能から逆算 https:// fed.brid.gy/r/https://gigazine .net/news/20260520-gemini-3-5-flash-parameter/

  4. Mastodon — mastodon.social TIER_1 (CA) · [email protected] ·

    llm-gemini 0.32 https://simonwillison.net/2026/May/19/llm-gemini-2/#atom-everything # AI # LLM # OpenSource

    llm-gemini 0.32 https://simonwillison.net/2026/May/19/llm-gemini-2/#atom-everything # AI # LLM # OpenSource

  5. Mastodon — mastodon.social TIER_1 (CY) · [email protected] ·

    llm-gemini 0.32a0 https://simonwillison.net/2026/May/19/llm-gemini/#atom-everything # AI # LLM # Python

    llm-gemini 0.32a0 https://simonwillison.net/2026/May/19/llm-gemini/#atom-everything # AI # LLM # Python