Bluesky Thread

GPT-5.1: a personality upgrade

November 13, 2025 View original thread

GPT-5.1: a personality upgrade

Instant is now a better conversation partner, and Thinking applies thinking more dynamically

* 2x fast for easy requests
* 2x slow for hard requests

openai.com/index/gpt-5-1/

A black-background infographic comparing GPT-5 (Standard) and GPT-5.1 (Standard) on how many tokens they generate for tasks of varying difficulty.

Title: “GPT-5.1 spends less time on easy tasks, and more time on hard tasks”

At the top are two legend dots:
• Light pink: GPT-5 (Standard)
• Bright pink: GPT-5.1 (Standard)

Below is a horizontal bar chart showing # of model-generated tokens per response across percentiles of task difficulty:
• 10th percentile: GPT-5 bar is small; GPT-5.1 bar is much smaller, labeled –57%
• 30th percentile: GPT-5 bar medium; GPT-5.1 bar shorter, labeled –31%
• 50th percentile: two nearly equal bars, labeled –0%
• 70th percentile: GPT-5.1 bar taller than GPT-5, labeled +21%
• 90th percentile: GPT-5.1 bar much taller, labeled +71%

At the bottom is explanatory text:

“GPT-5.1 Thinking varies its thinking time more dynamically than GPT-5 Thinking. On a representative distribution of ChatGPT tasks, GPT-5.1 Thinking is roughly twice as fast on the fastest tasks and twice as slow on the slowest tasks. Thinking time was set to Standard for both models.”

23 1

This definitely addresses a big complaint of mine — GPT-5-Thinking is very slow, but Instant isn’t trustworthy

But it’s also a bit of a letdown given what seems to be coming down the pipes from GDM

notable:

- it’s more efficient
- they removed GPT-5 for free users

feels like they’re moving into a new phase

6 1

More like this