01What is Claude Opus 5 and How It Works

Claude Opus 5 is a new AI model from Anthropic that performs at the same level as Fable 5 but costs half as much. It is built for agentic terminal coding, knowledge work, and computer use. In this episode, Julian Goldie tests it across many builds and shows that it holds its own against top models.

The model works by understanding prompts and generating code, text, or actions. It can create games, automate tasks, and handle complex reasoning. It is designed to be cheaper and more accessible than Fable 5, which is limited to 50% of your subscription on Claude.

With Opus 5, you get more usage for less cost, making it a practical choice for many users.

02Key Features and Benchmark Scores

Claude Opus 5 has several key features that make it stand out. It is better than Opus 4.8 in coding, reasoning, and real-world tasks. The model also has a fast mode that runs 2.5 times faster, though it uses more tokens and costs more.

  • Coding: scores over 2x Opus 4.8 on Frontier Bench.
  • Reasoning: scores about 3x the next best model on ARC AGI-3.
  • Real work: passes about 1.5x the rate of the next competitor on Zapier's automation bench.
  • Churn prevention: hits 100% on end-to-end workflow.
  • Safety: safer than previous models.
  • Computer use: better than Opus 4.8.

These scores show that Opus 5 is powerful and efficient. Julian emphasizes that you should test it yourself because benchmarks can differ from real use.

03How to Get Started with Claude Opus 5

To get the most out of Claude Opus 5, Julian recommends using it inside the AgentOS system. This system lets you access the model through a CLI without paying for API tokens, and you can plug it into various workflows.

  1. Sign up for the Airprofit Boarding at theairprofitboarding.com.
  2. Download the AgentOS zip file and install it.
  3. Set up the Obsidian memory system to give Opus 5 instant context.
  4. Access Opus 5 through the chat or CLI interface.
  5. Start building projects, like games or SEO workflows, with one-click access.

This setup allows you to use Opus 5 alongside other models like Hermes, Antigravity, and Codex. You can run parallel jobs and manage everything from one dashboard.

04Real Test: Building Games and More with Opus 5

Julian tested Opus 5 by creating several games in one shot. These builds show the model's ability to handle physics, graphics, and detail. For example, a driving game looked and felt competitive with Fable 5.

A 3D snake game was less interesting and not as fun to play, showing some inconsistency.

Other builds included a 3D racing game, an open-world RPG, and a flying simulator. The racing game was better than most frontier models. The RPG had nice 3D graphics and a moving world.

The flying simulator held its own, but controls felt weird. One flight simulator didn't work at all, but Julian notes these are one-shot builds and can be refined.

Overall, the model creates impressive stuff. Julian says it's right up there with Fable 5 on many tasks, sometimes even better.

05Results and Comparisons: Opus 5 vs. Other Models

Julian compared Opus 5 to Fable 5, GPT 5.6 Sol, and Kimi K3. On a driving game, Opus 5 matched Fable 5 in detail and quality. On another game, GPT 5.6 felt weird in colors, and Fable 5 had less nice controls.

Kimi K3's version was not bad but boring.

On CursorBench, Opus 5 lands within 0.5% of Fable 5. It performs better on many benchmarks, and it is cheaper. Julian says it's a no-brainer to switch from Fable 5, especially since you get more tokens with Opus 5 on your Claude subscription.

The model also tops automation tests and hits 100% on churn prevention. It scores 3x the next model on ARC AGI-3. These results show that Opus 5 is a strong competitor.

06Tips, Limits, and Mistakes to Avoid with Opus 5

Claude Opus 5 is not perfect. Julian notes that some builds have control issues, and one game didn't work at all. These are one-shot builds, so you can go back and forth to improve them.

Also, Opus 5 automatically routes to Opus 4.8 sometimes, just like Fable 5, which can be annoying.

  • Don't rely on theoretical benchmarks; test the model yourself.
  • Be prepared for occasional routing to older versions.
  • Expect some one-shot builds to have flaws; refine them.
  • Use fast mode carefully because it uses more tokens and costs more.
  • Pair Opus 5 with other models for fusion panels to get better outputs.

Julian emphasizes that the model is new, so patience is key. With time, it will only get better.

07Who Should Use Claude Opus 5 and Why

Claude Opus 5 is for anyone who wants powerful AI without the high cost of Fable 5. It's great for developers, creators, and businesses that need coding, automation, or creative builds. Julian says he is not technical and doesn't understand code, but he can create amazing stuff with it.

The model is especially useful for SEO workflows, game development, and computer use. With the AgentOS, you can plug Opus 5 into a system that manages all your AI agents, memory, and projects. This makes it easy to scale and stay ahead as models change.

Whether you're a beginner or expert, Opus 5 offers a cost-effective way to leverage cutting-edge AI. Julian recommends getting the AgentOS to get the most out of it.