Перейти к содержимому

Gemini Argon announced : here is everything you need to know

echohive

0:00 / 0:00

Gemini Argon announced : here is everything you need to know

1 399 просмотров · Трансляция закончилась 15 часов назад
echohive
12,6 тыс. подписчиков
1 399 просмотров · Трансляция закончилась 15 часов назад
AI keeps opening new possibilities. Let’s make them useful to you. PRIVATE CONSULTING Your project, AI workflow, or next big decision, let’s work through it together. You don’t need a perfectly formed question to begin. https://www.echohive.ai/consulting GET AMPLIFIED An evolving field guide to building with AI, making sense of markets, and exploring the mind. https://www.patreon.com/collection/18... 1000x LAB Explore what’s new, exchange ideas, and bring your questions. Live Sundays, with replays.   / 759209   Chapters 0:00 Gemini 4 Argon: Announced, Not Yet Yours 1:14 Argon's Price: $2 Now, $4 Later 2:29 Argon's Benchmarks: Where It Wins, Loses 3:51 Argon's One Million Token Output Limit 5:08 Test Your Agent Against Hidden Text 6:26 Hacker News Reacts to Argon 7:45 Argon on Google's Own Code: Rust, Memory, Video 8:59 Inside Fairwind: Who Gets Argon 10:30 Gemini 4 Argon: Where It Wins, Where It Lags 11:49 Gemini 4 Argon: Ahead or Just Level? 13:06 Reading Argon's Benchmarks With a Skeptical Eye 14:20 Leaked Argon vs. Launched Argon 15:36 Argon: Number One or Number Eight? 16:47 Google's Four Safety Gates Before Argon Ships 18:03 Argon's 95% Cache Discount Changes the Math 19:47 Argon's 40% Quantum Claim: Check the Receipts 21:08 Why Argon Skipped Gemini 3.5 Pro's Name 22:23 Argon's Eyes: Video, Charts and Graphs 23:41 Gemini 3.8 Flash: The Gemini You Can Use 24:54 Why Argon Waits on Washington 26:11 Argon's Vals Index Lead Is Under Two Points 27:27 Count Argon's Benchmark Wins Yourself 28:39 Did One Headline Really Move Alphabet? 29:59 Terminal-Bench 4.0 Got Harder: Read Scores Right 31:16 Argon's DeepSWE Win, Measured in Points 32:38 Harvey Legal Benchmark: 19.6% Is the Win 33:56 Argon's Hype Demos: One Sample Isn't Proof 35:13 Argon's Cyber Score: Check the Baseline 36:34 Argon: Weird Name, Hard to Find? 37:44 Argon's Benchmarks Aren't Apples-to-Apples 39:38 GraphWalks: A Big Lead, A Small Test 40:56 Argon's Finance Win: Check the Real Leaderboard 42:12 Argon Landed in OpenAI's Rough Week 43:32 Outside Score: Argon Ties GPT-6 Astra 44:47 Argon Hallucinates Less: 15% vs 51% 45:59 Argon Is Verbose: Watch the Token Bill 47:16 What Google Hasn't Documented About Argon 48:36 Was Gemini Cooked Before Argon? 49:56 Argon's 91.9% Is Second on Vibe Code Bench 51:10 Argon's Long-Context Cliff: 99.7% to 84.2% 52:31 Argon Ranks 32nd: Check the Coverage 53:44 Alphabet After Hours: Up 1% or 2%? 55:05 Why Top Labs Gate Their Cyber Models 56:28 Same Test, Different Scores: Terminal-Bench 4.0 57:47 Independent Testers Still Rank Opus 5.5 First 59:35 Cheapest Per Token Isn't Cheapest Per Task 1:00:52 Google Won't Train on Its Own Safety Findings 1:02:06 Argon Also Trails Fable 5.1 on Two Tests 1:03:22 Check the Effort Label Before the Score 1:04:36 The 5x Price Gap That Shrinks 1:05:48 Fairwind: 650 Organizations or Undisclosed? 1:07:01 Argon's Video Score vs Image-Only Listing 1:08:17 Your Eval Beats Every Launch Table 1:09:32 One Model, Two Tiers of Access 1:10:50 Is Gemini 4 Argon Really Thrifty? 1:12:02 AutomationBench: What Each Task Really Costs 1:13:20 Google Asks the Industry to Keep Reasoning Visible 1:14:35 Argon: First Flagship Under New DeepMind Leaders 1:15:53 What One Million Output Tokens Actually Costs 1:17:13 Keep Your Pro Setup Until You Compare 1:18:58 Argon's Top Coding Score Has Shaky Grading 1:20:12 Argon's Weakest Test Has Gameable Grading 1:21:27 Is Gemini 4 Argon Just a First Draft? 1:22:43 Will Gemini 4 Argon Keep Its Name? 1:23:59 The Argon Thread Praised Flash Instead 1:25:21 Argon's Price Isn't a Discount 1:26:38 The White House AI Accord, One Day Earlier 1:27:59 1M Output Tokens: 16x, Not Quadrupling 1:29:22 Beware Confident Argon Architecture Explainers