Updated Friday August 8, 5:21 pm ET: Shortly after this post’s publication, OpenAI co-founder and CEO Sam Altman announced the company would restore access to GPT-4o and other previous models for certain users, acknowledging that the GPT-5 launch was “more challenging than we hoped for.”

The much-anticipated release of OpenAI’s new model, GPT-5 is experiencing a rocky beginning, to say the least.

Despite overlooking errors in charts and voice demos during yesterday’s livestreamed presentation of the new model (which actually consists of four distinct models and a ‘Thinking’ mode applicable to three of them), a number of user reports have surfaced since GPT-5’s release, highlighting significant errors in solving relatively straightforward problems that prior OpenAI models — and competitors from other AI labs — correctly addressed.

For example, data scientist Colin Fraser shared screenshots demonstrating GPT-5 incorrectly solving a math proof (whether 8.888 repeating equals 9 — it certainly does not).

AI Scaling Hits Its Limits

Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:

• Turning energy into a strategic advantage

• Architecting efficient inference for real throughput gains

• Unlocking competitive ROI with sustainable AI systems

Secure your spot to stay ahead: https://bit.ly/4mwGngO

It also struggled with a basic algebra arithmetic problem that elementary schoolers could likely solve, 5.9 = x + 5.11.

Using GPT-5 to evaluate OpenAI’s flawed presentation charts didn’t result in helpful or accurate responses.

It also stumbled on this more complex math word problem below (which, to be fair, initially puzzled this human…though Elon Musk’s Grok 4 AI solved it accurately. For a hint, consider that flagstones in this scenario cannot be divided into smaller pieces. They must remain whole as 80 separate units, so no halves or quarters).