GPT-5.1 fails to deliver major improvements over GPT-5, tests show

OpenAI’s latest multimodal model, GPT-5.1, has arrived — but early testing suggests it offers only marginal improvements over GPT-5 despite the company’s claims of enhanced tone, speed, and customisation.

According to a review by PCMag, real-world use shows little difference in response quality or performance, despite OpenAI promoting GPT-5.1 as “faster and more intelligent.” The expected boost in human-like interaction has also fallen short.

OpenAI has introduced new built-in personality presets — Friendly, Efficient, Candid, Professional, and Quirky — replacing earlier options. While ChatGPT’s behaviour can still be customised, testers found that the output feels only slightly more refined.

The model does show some subtle tonal shifts. For example, GPT-5.1 offered a warmer and more empathetic response than GPT-5 when asked about anxiety. However, this increased warmth wasn’t consistent across topics; discussions on gaming produced nearly identical responses between the two versions.

Critics note that despite OpenAI’s promises of improved personalisation, GPT-5.1 does not make ChatGPT noticeably more human-like than GPT-5. Overall, the update delivers minor enhancements rather than the transformative leap many users anticipated.