Wes Roth

Grok 4.5 just COOKED Claude and OpenAI

Item: Grok 4.5
Author: Wes Roth

Jul 9, 2026 35 min

ai modelsgrok 4.5xaitechnologyartificial intelligence

Watch on YouTube Follow Wes Roth on Rundown — free

Summary

AI summaries can be incomplete or wrong. Verify anything important against the original video.

The video introduces Grok 4.5, a new AI model from xAI, highlighting its speed, cost-effectiveness, and capabilities in coding, agentic tasks, knowledge work, and office tasks. It compares Grok 4.5's performance against other leading models like Opus and GPT on various benchmarks and showcases its features through demos.

The video begins by introducing Grok 4.5, emphasizing its speed (80 TPS), cost-effectiveness, and improvements over previous models, including its ability to handle complex Excel formulas, research from the web, and even draft PowerPoint presentations. The presenter then dives into benchmarks, showcasing Grok 4.5's performance against Opus 4.8 and GPT 5.5 in various categories like DeepSWE 1.0, DeepSWE 1.1, Terminal Bench 2.1, and SWE Bench Pro resolve rate. Grok 4.5 consistently performs well, often outperforming or matching competitors in these benchmarks. The video then shifts to demonstrating Grok 4.5's capabilities through several demos: creating a 3D ship sailing game with voice commands, designing a CRM dashboard for a small business, generating SVG illustrations, and automating Google Sheets workflows. A significant portion is dedicated to the "Living City" demo, showcasing the AI's ability to generate intricate 3D environments. The presenter expresses enthusiasm for the model's performance, especially its creative generation capabilities, despite noting some minor issues like clipping in the 3D city demo. The video also briefly touches on the pricing, highlighting its competitive edge. Finally, the presenter mentions the potential for Grok 4.5 to be a game-changer for AI workflows, enabling users to build custom dashboards, CRMs, and workflows with ease. The video concludes by showing the results of various prompts, including 3D scenes and illustrations, underscoring the model's versatility.

Verdict

Grok 4.5

AI model · $2 per million input tokens and $6 per million output tokens

Grok 4.5 represents a significant advancement in AI models, offering impressive performance, versatility, and cost-efficiency across various tasks, from coding to creative generation.

Buy

Pros

✓ Fast and low cost 2:10
✓ Significant step up over previous models 3:20
✓ Excels at office tasks (Excel, research, PowerPoint) 2:30
✓ High performance in benchmarks 8:10
✓ Capable of generating complex 3D environments
✓ Competitive pricing

Cons

✗ Minor visual glitches like clipping in 3D demos

Specs

TPS	80	1:40
Token Efficiency	4.2x fewer tokens than Opus 4.8	6:20
Cost (input tokens)	$2 per million
Cost (output tokens)	$6 per million

Compared to

Opus 4.8

Grok 4.5 has significantly higher token efficiency and is cheaper.
GPT 5.5 (High)

Similar performance in speed, but Grok 4.5 is cheaper.

Best for

Not for

Claims & arguments

Grok 4.5 performance

Grok 4.5 offers significant improvements in speed, cost-efficiency, and performance across various tasks compared to previous models and competitors.
- 2:10 80 TPS fast-model speeds and twice greater token efficiency than latest leading models.
- Significantly cheaper than Opus 4.8 based on benchmark costs.
- 8:10 Performs competitively or better than Opus 4.8 and GPT 5.5 in multiple benchmarks (DeepSWE, Terminal Bench, SWE Bench Pro).
Grok 4.5 capabilities

Grok 4.5 is highly versatile and capable of handling complex tasks in coding, office work, and creative generation.
- 2:30 Excels at coding, agentic tasks, knowledge work, complex Excel formulas, web research, and PowerPoint creation.
- Can generate detailed 3D environments and scenes.
- Capable of creating SVG illustrations, automating workflows, and generating text.
Grok 4.5 cost and availability

Grok 4.5 is competitively priced and widely available, making advanced AI accessible.
- Offered at a competitive cost compared to other leading models.
- Available in Grok Build, Cursor, and SpaceX AI console.

Key Points

2:30 Grok 4.5 excels at coding, agentic tasks, knowledge work, and office tasks like Excel formulas, web research, and PowerPoint creation.
8:10 Benchmark comparisons show Grok 4.5 outperforming or matching competitors like Opus 4.8 and GPT 5.5 in various tests.
35:00 Demonstration of Grok 4.5 creating a 3D ship sailing game with voice commands.
Introduction of Grok 4.5 by xAI, emphasizing speed, cost-effectiveness, and improved performance over previous models.
Demonstration of Grok 4.5 designing a CRM dashboard for a small business.
Demonstration of Grok 4.5 generating SVG illustrations and automating Google Sheets workflows.
Showcase of 'The Living City' demo, highlighting Grok 4.5's ability to generate complex, detailed 3D environments.
Discussion of Grok 4.5's pricing, noted as competitive compared to other leading models.
Mention of Grok 4.5 being available in Grok Build, Cursor, and the SpaceX AI console.
The model's potential to revolutionize AI workflows by enabling users to build custom dashboards, CRMs, and workflows.
Brief look at other generated content including images and a summary of a Twitter post comparing different AI models.
Comparison of Grok 4.5's performance with Fable 5, highlighting Grok's efficiency and cost-effectiveness.

Worth watching if: This video is worth watching if you are interested in the latest advancements in AI models, specifically comparing the performance and capabilities of Grok 4.5 against other leading models like Opus and GPT, and seeing practical demonstrations of its applications.

Get every Wes Roth video extracted like this

One daily email with structured extracts of every channel you follow. Free tier covers 15 videos a month.

No credit card. Free tier forever.

Watch on YouTube