Sponsored by
n
Anthropic just dropped Claude Sonnet 5, and the marketing says "most agentic Sonnet yet."
The token price stayed flat at $2 per million input tokens.
But here's the twist: real-world costs are climbing because the model burns through way more tokens to get the job done.
But wait, there's more:
Google shipped two new media models that let you chain image generation straight into video workflows
Anthropic launched Claude Science, a research workbench that plugs their models into lab tools and scientific databases
I'm Alex. Welcome to L8R by Innov8.
Let's dive deep 🐰
In today's post:
Claude Sonnet 5 costs more than you think
Google's new models turn images into videos in seconds
Anthropic built a lab assistant for scientists
Anthropic's Sneaky Price Hike with Claude Sonnet 5
Anthropic just dropped Claude Sonnet 5, calling it their smartest mid-tier model yet.
But do not let the sticker price fool you.
A sneaky change to how it reads text means you will pay a lot more to use it.
The details:
Anthropic released Claude Sonnet 5 on June 30, 2026, with a massive 1-million token context window.
The model uses a new tokenizer that eats up about 30% more tokens for the exact same text compared to Sonnet 4.6.
It starts at $2 per million input tokens, but that price jumps to $3 on August 31, 2026.
You cannot change settings like temperature anymore; the API will just give you an error.
The model talks a lot more to solve problems, which drives your real-world costs even higher.
Why it matters:
Businesses look at the base token price and think their costs will stay flat. They are wrong.
Between the hungry new tokenizer and the model's extra thinking steps, your actual bill will shoot up.
You might end up paying more for Sonnet 5 than you did for the giant Opus 4.8 model on certain tasks.
💡 L8R's Take:
Anthropic is playing games with their pricing.
Hiding a price hike behind a new tokenizer is a cheap trick.
If you build with Claude, you need to watch your API bills like a hawk this month..
From Our Partner
37 Free Claude Prompts With The AI Report
Subscribe to The AI Report, the free 5-minute daily AI brief for 400,000+ business leaders, and you’ll get 37 Claude prompts free in your welcome email. They’re organised by the 8 situations every manager faces. You get both: the newsletter and the prompts.
Google Drops Fast Images and Chatty Video AI
Google just dropped two wild new AI models for developers.
You can now build apps that pump out super cheap images and let you edit videos just by chatting.
This changes the game for anyone making content at scale.
The details:
Nano Banana 2 Lite spits out 1k resolution images in under four seconds.
It costs practically nothing at just $0.034 for every 1,000 images you make.
Gemini Omni Flash is a new video model that lets you edit clips by talking to it in plain English.
The video model costs about $0.10 per second and maxes out at 10-second, 720p clips right now.
Google released both models on June 30, 2026, letting developers chain them together to turn images into videos instantly.
Why it matters:
Making tons of content used to take huge amounts of time and money.
Now, a developer can build a tool that generates a product photo and turns it into a video ad in seconds.
Google wants to own the entire creative pipeline. They are making it too cheap and easy to ignore.
💡 L8R's Take:
Nano Banana 2 Lite is the real star here because of that insane price tag.
Gemini Omni Flash is a fun toy, but a 10-second limit at 720p is still pretty weak for pro video.
Either way, chaining these two together will flood the internet with AI ads faster than we can blink.
From Our Partners
How Much Is Your Billing Lag Actually Costing You?
Most SaaS finance teams know their billing process isn't perfect. Few know what it's actually costing them.
Answer 5 quick questions — contracts signed per month, ACV, days to first invoice, error rate, DSO — and the Tabs Billing Lag Calculator gives you a dollar figure benchmarked against top SaaS companies.
It takes two minutes. The number might surprise you.
Calculate your billing lag and see where you stand.
Anthropic Builds a Lab Bench for AI Scientists
Anthropic just dropped a massive upgrade for researchers called Claude Science.
It is not a new AI model, but a smart workspace that links Claude to heavy-duty research tools and databases.
This aims to fix the messy, broken workflows that slow down drug discovery and biological research.
The details:
Claude Science connects existing Claude models to over 60 scientific databases and tools like PubMed, Jupyter, and R.
It partners with NVIDIA to access powerful biological AI models like Evo 2 and OpenFold3.
A built-in reviewer agent double-checks citations and math to stop AI hallucinations in their tracks.
Every chart or paper it makes comes with a full, traceable history of the code and data used to build it.
Anthropic is even using this new tool to start its own drug discovery program for neglected diseases.
Why it matters:
Scientists waste way too much time jumping between different software, terminals, and notebooks.
Claude Science puts everything in one place so researchers can focus on actual science.
It also kicks off a massive battle with OpenAI, who launched their own science AI called GPT-Rosalind in April.
Whoever wins this race will shape the future of medicine.
💡 L8R's Take:
I love seeing AI move past simple chatbots and into actual, messy human workflows.
Anthropic building a tool that proves its own math with a clear audit trail is exactly what the industry needs right now.
If this actually speeds up finding new cures, it is the most important AI launch of the year.
🚀 Quick L8R Summary
Claude Sonnet 5: Anthropic's new mid-tier model does more work per task, so your bills go up even though token prices stayed flat.
Google's Nano Banana Lite & Gemini Omni Flash: Two new models that let developers chain image generation straight into video editing workflows.
Claude Science: Anthropic built a research workbench that plugs Claude into lab tools and scientific databases for drug discovery teams.
📩 Innathe L8R engane undarunnu 👇?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bie…👋