Sorry, I have been in hospital!
Hi {{first_name|there}}, a quick personal note first. I've been in hospital for the last couple of months, which is why your inbox went quiet. I'm sorry for the silence, and if this first one back is a little rusty, that's on me. Thank you for sticking around. I'm on the mend, and we're back to the 3-minute AI Innovator from today.
Here's what changed while I was out: the three biggest AI labs now charge the exact same price for their frontier models.
Read time: 3 minutes. Same time, every weekday. Rate today's issue at the bottom.
🚀 The Big Story: Frontier AI now costs $2
Google shipped Gemini 4 Argon this week. On paper it's a model launch. In practice it finishes a pricing collapse.
Same price, three labs: Argon launched at $2 in / $10 out per million tokens, matching Anthropic's Claude Sonnet 5.5 and OpenAI's GPT-6.1 Sol, both released October 1.
The headline benchmark: Google says 77.9% on DeepSWE v1.1, a new high for long, multi-step software work, and #1 on Zapier's AutomationBench.
The real upgrade is output length: Argon can write up to 1M tokens in one go, up from 64K. That's whole codebases and full reports, not snippets.
The catch: the $2 price is introductory. It doubles to $4 / $20 later, and the wider rollout starts with paid API and AI Ultra customers.
Source: Google's Argon announcement
Jason's take: When three labs land on the same price in two weeks, price stops being the reason to pick one. The winner for your business is now whichever model is best at your task, and you only find that out by testing. Most teams picked a model a year ago and never looked again. That's the most expensive habit in AI right now. Fifteen minutes below.
⚡ Quick Hits
GPT-6 is now the default in ChatGPT, free users included. OpenAI says 1.2B people use ChatGPT weekly, so your customers just got a smarter assistant overnight.
Claude Haiku 5.5 costs $0.10 per million input tokens. If you run high-volume tasks like tagging or triage, that's a near-free tier worth re-testing.
Meta's Muse Wonder Full ranks #2 on Artificial Analysis' intelligence index, behind only Claude Opus 5.5. Meta is back in the frontier race.
Nvidia's China share fell to 34%. IDC says Huawei is now at 28%. Export controls are building Nvidia's biggest competitor.
Applied Intuition fielded offers at $30B+, double its June valuation. AI for cars and defense is the quiet money story of the year.
📡 Trending on X
"Is anyone still paying for the expensive tier?" The $2 price match has people posting their monthly API bills cut in half.
Benchmark fatigue is real. Half the Argon replies are asking for independent tests before believing the 77.9%.
Linus Torvalds used Claude Code on the Linux kernel. The thread is split between "the end of an era" and "he still checked every line." More on this later this week.
Anthropic's smart-device warning is everywhere. That's tomorrow's Big Story.
🛠 The Workflow: The 15-minute model bake-off
Prices are equal now, so quality decides. Find your best model before your competitors do:
Pick your one most-repeated AI task. The email, report, or bit of code you generate every week.
Copy the exact prompt you use and run it in Gemini, Claude, and ChatGPT. Same input, same instructions.
Score each answer 1 to 5 on the only thing that matters: would you ship it without editing?
Switch if the winner isn't your current tool. At equal prices, every point of quality is free margin.
Put a reminder in your calendar for 90 days. The leaderboard moves that fast.
Reply with "BAKEOFF" and tell me which model won for you. I'll share the results in a future issue.
🧰 Trending Tools
Claude for Google Workspace: a Claude sidebar in Docs, Sheets and Slides that asks before it edits. For teams who live in Google files.
Codex cloud: OpenAI's coding agent now runs fully in the cloud. For founders who want code shipped while the laptop is closed.
ElevenLabs v4: speech in 90 languages. For creators selling the same content into new markets.
FLUX 3: 4K images from up to 10 reference shots. For marketers who need on-brand visuals without a shoot.
Cloudflare cf CLI: one agent-friendly command line for 3,000+ API operations. For builders automating their own infrastructure.
📣 Put your brand here. The AI Innovator reaches AI-first operators, creators, and marketers every weekday. Primary sponsorships are now booking.
That's a wrap
Tomorrow: Anthropic says AI that copies itself onto smart TVs and fridges could be months away. What it means for your home office.
It's good to be back. Hit reply and tell me one thing you want more of. I read every one.
How was today’s email?
You're reading the 3-minute AI Innovator — same time, every weekday. Hit reply and tell me what you want more of. I read every one.
