OpenAI should be arrested for this

@avelishealth

3.8K followers · usually 5.4K views a post

Views

30.0K

Likes

957

Comments

68

Shares

71

SoundTikTok sound · original sound - avelishealth

Why it worked

The video leverages a controversial statement in the on-screen text and a potentially misleading chart from a well-known company (OpenAI) to generate engagement and discussion in the comments.

Summary

The creator questions a chart posted by OpenAI comparing GPT-5 to older models, suggesting there's something wrong with the data presented. The chart appears to be a benchmark for software engineering tests.

Structure

1Introduction of the chart and claim
2Explanation of the chart's purpose
3Questioning the chart's validity
4Call to action to identify errors

Product placement

The video discusses and displays a chart from OpenAI, specifically mentioning GPT-5, GPT-4o, and OpenAI o3, in the context of a software engineering benchmark.

Call to action

Can you see anything wrong with this graph?

On-screen text

OpenAI should be arrested for this SWE-bench Verified Software engineering Without thinking With thinking GPT-5 74.9 52.8 OpenAI o3 69.1 GPT-4o 30.8 10:16 AM - 8/7/25 766K Views

Transcript

Original caption

I just wanna know who at OpenAI made this chart #openai #chatgpt #ai #Tech #gpt5

More from @avelishealth

More breakout videos from this creator.

I'm just gonna put this here. I want you guys to see if you can find anything that is wrong with this chart. OpenAI posted this by the way on their actual live stream for GPT5. This is the benchmark for the software engineering test. So obviously they want to make GPT5 seem like it's the best model yet, right? Can you see anything wrong with this graph?