Google's Gemini 3.5 Pro Is Finally Here — The 2-Million Token Monster That Had To Be Rebuilt From Scratch

-

I remember when everyone said Google was falling behind in AI. Well, today they just dropped Gemini 3.5 Pro, and I have to say — this thing is not what I expected. Not because it's bad. Because it's actually good. Really good. And the story behind how it got here is even more fascinating than the model itself.

They Scrapped the Original Version and Started Over

Here's the part that blew my mind: Google's engineering team discovered structural failures in the original Gemini 3.5 Pro — specifically in how it handled recursive tool-calling, the technique where an AI calls one tool, then uses the result to call another, chaining actions together. The failures were bad enough that Google made the extraordinary decision to scrap the base model entirely and restart pretraining from scratch. That's like demolishing a skyscraper that's already 60 floors up because the foundation is cracked. It's painful, expensive, and it delayed the release by months. But it shows Google's engineers aren't willing to ship garbage just to hit a deadline.

What Gemini 3.5 Pro Actually Brings to the Table

Let's talk specs. Gemini 3.5 Pro launches with a 2-million-token context window — that's roughly 1,500 pages of text you can feed it in a single conversation. For context, GPT-4o tops out at 128K tokens. A 2M context window means you can hand this model an entire codebase, a year's worth of emails, or a feature-length screenplay and ask it to reason across all of it at once.

The model also features a "Deep Think" reasoning mode, available exclusively on Google's $250/month Ultra tier. This is Google's answer to OpenAI's o3 and Anthropic's extended thinking — a slow, deliberate reasoning mode designed for problems that need to be chewed on rather than answered instantly. Pricing for the API lands at approximately $1.25 per million input tokens and $10 per million output tokens, putting it competitive with Claude Sonnet 4 and slightly below GPT-4o.

The Coding Problem They Couldn't Ignore

Even with the restart, Gemini 3.5 Pro had a rough road to today's launch. Internal testing revealed the model fell short in two critical areas: coding performance and complex long-horizon reasoning tasks — exactly the things enterprise customers care most about. The model stayed in limited enterprise preview while Google's team worked to close the gap. According to multiple sources, it remained there until engineers were satisfied with benchmark parity against competitors.

The fact that Google is launching today, July 17, suggests they believe they've finally crossed that bar. Whether the benchmarks translate to real-world performance is something the developer community will figure out in the next 48 hours.

Why This Launch Matters

Google being competitive in the frontier model race isn't just good for Google — it's good for everyone. Competition between Gemini, Claude, GPT-4o, and open models like Kimi K3 (also released this week by China's Moonshot AI) means prices drop and capabilities improve. When Anthropic sees Google offering a 2M context window at $1.25/M input tokens, they respond. When OpenAI sees Deep Think competing with o3, they push harder on reasoning.

The real winner here is developers and enterprises who get access to increasingly powerful AI tools at prices that were unimaginable just 18 months ago.

My Take

I've been following the Google AI story since Bard launched to mixed reviews in early 2023. The journey from that stumbling start to a model with a 2M context window that had to be rebuilt from scratch just to meet quality standards — that's a remarkable transformation. Google had to earn this moment through some genuinely painful decisions. I'm curious whether Gemini 3.5 Pro will convert skeptics, or whether the developer ecosystem has already committed too deeply to Claude and GPT-4.

The next few weeks of benchmarks and real-world testing will tell us everything.

What's your experience? Drop a comment below! 👇

Have you tried Gemini 3.5 Pro yet? What use case are you most excited to test it on — the 2M context window, the Deep Think mode, or something else entirely?

Comments

Post a Comment

Popular posts from this blog

This AI Startup Is Worth $26 Billion and Writes 90% of Its Own Code — Should Software Engineers Be Worried?

-
I've been watching the AI coding space very closely for the past year, and every few months something happens that forces me to recalibrate my sense of how fast things are moving. This week, that moment came from Cognition AI — the startup behind Devin, the AI software engineer — which just announced a $1 billion funding round at a $26 billion valuation . But the number that actually stopped me cold wasn't the valuation. It was this: 90% of Cognition's own code is now written by Devin. From $10B to $26B in Eight Months — The Numbers Are Staggering Let's put this in perspective. Just eight months ago, in September 2025, Cognition closed a $400 million round at a $10.2 billion valuation. Today, that valuation has more than doubled to $26 billion. The new round was co-led by Lux Capital, General Catalyst, and 8VC, with Ribbit Capital, Atreides Management, and Peter Thiel's Founders Fund also participating. But what's really driving this isn't hype — it...
Read more

Sony Smart Tags Review: The NFC Trick That Made My Life 10x More Convenient (Before Everyone Knew NFC Existed)

-
In 2012, NFC (Near Field Communication) was a technology that most smartphone owners had never used or even heard of. Sony released Smart Tags — small NFC stickers you could program to trigger actions on your Android phone with a single tap — and I became obsessed with them. Here's why they were brilliant, and why they pointed toward a future most people are only now catching up to. What Are Sony Smart Tags? Sony Smart Tags are small NFC-enabled stickers (about the size of a large coin) that you program using the NFC Smart Tags app on Android. Each tag can be loaded with a set of actions that trigger automatically when your phone touches it. One tap on the tag = multiple phone settings change instantly. Sony sold them in packs of 4, with different colors and designs. You stuck them places — your nightstand, your car dashboard, your desk, your bag — and each location became a context-aware shortcut. How I Actually Used Them Bedside tag (white): One tap when I got into bed tr...
Read more

WWDC 2026 Preview: Apple Needs to Fix Siri or It's Game Over for Apple Intelligence

-
Apple's WWDC 2026 kicks off on June 8, and the stakes couldn't be higher. While Google announced 900 million Gemini users at I/O last week, Apple is heading into its biggest developer conference with one uncomfortable question hanging over the whole event: what happened to Siri? I've been using Apple devices daily since the original iPhone, and I'll say it plainly — Apple Intelligence has been the most underwhelming software launch in Apple's recent history. WWDC 2026 is Apple's chance to fix that. Here's what they need to announce, what we're expecting, and what a failure would mean. The Problem Apple Needs to Solve When Apple announced Apple Intelligence at WWDC 2024, the demos were impressive. When the features started rolling out through 2025, the reality was a different story. The "new Siri" was slower, less reliable, and often less capable than the old one for basic tasks. The "Personal Context" features that were supposed ...
Read more