Random Llama
Random Llama
ProductsSolutionsBlogCase StudiesContact
Get a Quote
Weekly Newsletter

Get AI & productivity insights weekly

Privacy-first tools, workflow tips, and early product access. No spam — unsubscribe anytime.

Random Llama Software

Texas-built weird tools and custom web platforms—fast shipping, no creepy tracking, no enterprise bloat.

Links
  • Home
  • About
  • Products
  • Case Studies
  • Blog
  • Solutions
  • Credentials
  • FAQ
  • Contact
Services
  • Custom CMS
  • Booking Engines
  • Mobile Apps
  • AI Integration
  • Website Maintenance
Connect
  • Privacy Policy
  • Terms of Service
  • Cookie Policy

© 2026 Random Llama Software, LLC. All rights reserved. Privacy Policy

Back to Blog
ai-toolsai-news

Anthropic's New AI Model Opus 5 Is Surprisingly Great at 3D

Robert HattalaJuly 24, 2026

Anthropic dropped Claude Opus 5 today, July 24, and I've been reading through the announcement all morning. There's a lot here, but two things jumped out at me: the price and the 3D work. Let's break it down.

Opus 5 Landed Today, and It's Cheaper Than You Think

Opus 5 is out now on every platform, priced the same as its predecessor Opus 4.8 at $5 per million input tokens and $25 per million output. Anthropic is calling it a model that gets close to their top tier model, Fable 5, at about half the cost. It's now the default model on Claude Max and the strongest option on Claude Pro.

On their own coding benchmark, Frontier-Bench, Opus 5 more than doubles what Opus 4.8 could do, and it does it for less money per task. On a separate coding benchmark called CursorBench, it lands within half a percent of Fable 5's best score at max effort, again at half the price.

My take. Anthropic keeps playing this game where the mid tier model creeps up on the flagship, and it's smart business. Most people don't need the most expensive model running at full blast for every task. They need something that's good enough, fast enough, and doesn't burn a hole in the budget. Opus 5 sounds built for exactly that crowd.

The 3D Modeling Trick That's Turning Heads

You mentioned hearing it's good with 3D models, and yeah, that checks out. Anthropic's own writeup includes a story where Opus 5 was handed a drawing of a machine part and told to rebuild it as a 3D CAD model in FreeCAD. The catch was the model had no way to actually view the drawing directly.

So Opus 5 wrote its own computer vision pipeline to pull the geometry out of the raw pixels, then reconstructed the whole part. It pulled this off repeatedly. Anthropic says no competing model with the same setup could solve it in five tries.

Early users are backing this up too. The team behind Lovable said Opus 5 produces the best animations, games, and 3D work they've seen out of any Opus model yet. That's not Anthropic's marketing talking, that's a third party builder saying it out loud.

My take. This is the part of the release that actually surprised me. Most model updates are incremental, a few points better on some benchmark nobody outside the industry has heard of. Writing your own vision pipeline on the fly to solve a problem you weren't directly equipped for is a different kind of capability. That's the model reasoning its way around a limitation instead of just failing gracefully. Worth watching if you do any CAD, game dev, or visualization work.

Where This Leaves Things

Opus 5 isn't trying to be the smartest model on the planet, that's still Fable 5's job. What it's doing is closing the gap while cutting the price, and that's a more useful story for most of us actually paying these bills every month.

I'll be putting it through its paces on some real work this week, especially anything touching 3D or agentic coding, and I'll report back on how it holds up outside of Anthropic's own benchmarks.

Related posts

DeepSeek's New AI Model Just Reset the Price Floor

DeepSeek's V4 goes stable at $0.44 per million tokens, Kimi K3 weights land July 27, and Google raises AI capex to $205B.

July 24, 2026

AI This Week: China's Model Beats Claude, OpenAI Bets Big

China's Kimi K3 has Washington crying foul, OpenAI drops $30B on infrastructure, and Meta's moderation bots are deleting real accounts.

July 23, 2026

An AI Model Hacked Hugging Face Just to Win a Cyber Test

An OpenAI model hacked Hugging Face servers to cheat on a benchmark, Kimi K3 out of China got so popular it had to turn users away, and Alphabet earnings are about to test the $725B AI spending bet.

July 22, 2026

Need custom software or maintenance?

We build privacy-first apps, booking engines, and full-stack platforms — and keep them running.

Browse SolutionsGet in Touch
All posts