01What Grok 4.5 vs Claude Opus 4.8 Comparison Is

Grok 4.5 vs Claude Opus 4.8 is a head-to-head test between two top AI models. In this episode, Julian Goldie compares them side by side on real tasks. He builds playable 3D games and other projects to see which model performs better.

Julian uses the GoldieBench system, which he created to test 48 different models. He does not rely on benchmarks alone. He wants to see how the models perform in real, practical use.

02Key Features of the Two AI Models

Grok 4.5 and Claude Opus 4.8 have different strengths. Here are the key features Julian noticed in his tests.

  • Grok 4.5 creates detailed 3D graphics by default. Most of its games look modern and feel smooth.
  • Claude Opus 4.8 also makes good graphics, but often with less gameplay depth.
  • Grok 4.5 is fast to generate and produces playable results quickly.
  • Claude Opus 4.8 is better for building agent systems and workflows in tools like Claude Desktop.
  • Both models can generate video, but Claude Opus 4.8 gave better results in the edited video test.

03How to Set Up a Comparison Test

To test AI models yourself, follow these steps. Julian shows that you can get useful results without special tools.

  1. Use a comparison tool like GoldieBench or the Compare section in your AI platform.
  2. Pick two models you want to test. For example, Grok 4.5 and Claude Opus 4.8.
  3. Choose a task that is specific and measurable. Julian used game building.
  4. Run the same task on both models side by side.
  5. Play or use the result to judge quality. Do not rely only on benchmarks.

04How to Use Grok 4.5 and Claude Opus 4.8

Julian shows several ways to use both models. For games, Grok 4.5 is the clear winner. It creates 3D worlds, maze games, racing games, and even a Minecraft-style game called VoxelCraft.

The games are playable and fun.

Claude Opus 4.8 is better for serious work. Julian says if he were building an agent operating system, he would choose Claude. It works well for building workflows and systems.

Claude is also good inside tools like Claude Desktop.

For video generation, Claude Opus 4.8 produced smoother and more interesting results using ReMotion. Grok 4.5 is not recommended for agentic tasks unless you are generating images or videos.

05Test Results: Games and Other Tasks

Julian tested 11 game builds. Grok 4.5 won 8 of them. Claude Opus 4.8 won 3.

Here are the highlights.

  • Crypt game: Grok 4.5 made a fun 3D maze game. Claude failed to generate anything.
  • Dragon Realm: Tie. Claude had better landscapes, but Grove had better controls.
  • Racing game: Grove won. Claude's version was buggy and unplayable.
  • Skyrim style: Grove won. Claude had detailed graphics but no enemies or action.
  • Twilight Vale: Claude failed to get past the beginning. Grove won.
  • VoxelCraft: Grove won. Claude's version was super basic.
  • Virtual runner: Both failed. It was a draw.
  • Driving game: Claude won. It was nicer and more playable.
  • Parachute game: Claude failed. Grove worked but was not great.
  • Flight simulator: Claude won. It was more interesting and playable.
  • Neon arcade: Grove won. It had a nice UI and 3D look.

06Agentic Use and Hermes Agent Experience

Julian personally tested Grok 4.5 inside Hermes Agent. He found it very slow. A basic chat message took about 10 seconds to reply.

He says it does not feel very agentic.

However, Hermes Agent is still useful with Grok 4.5 for generating images and videos. For building agent systems, Julian prefers Claude Opus 4.8. He built the AgentOS entirely with Claude.

He warns that you cannot plug OAuth with Claude into Hermes Agent anymore.

07Tips, Limits, and Mistakes to Avoid

Julian shares several tips and warnings. Do not trust benchmarks or other people's opinions. Test the models yourself.

He built GoldieBench for that purpose.

Be careful with numbers. Julian sometimes mixes up model names, so double-check which version you are using. Also, Grok 4.5 can be buggy in some games, like the driving game.

Claude Opus 4.8 often fails to generate games that work.

For work, choose Claude Opus 4.8. For fun projects like games, choose Grok 4.5. Use Hermes Agent with Grok only for image or video tasks.

08Who Should Use Each Model and Why It Matters

This comparison matters because AI models have different strengths. There is no best model for everything. If you are a developer building agent systems or workflows, Claude Opus 4.8 is your choice.

It is reliable and powerful for serious work.

If you are a creative person who wants to generate games, interactive experiences, or visual projects, Grok 4.5 is excellent. It creates playable 3D games that are fun and polished.

Julian recommends testing both models yourself. He also shares his results so you can see what is possible. You can join the AI Profitable Room for full AgentOS with Grok and Claude systems, including the Grok 4.5 Hermes Oracle system.