Bluesky Thread

terrible naming, should’ve called it gpt-4o-final(2)-large-lite

April 14, 2025 View original thread

terrible naming, should’ve called it gpt-4o-final(2)-large-lite

Tim Kellogg @timkellogg.me

GPT 4.1, 4.1-mini, and 4.1-nano

GPT-4.1 is cheaper & smarter than 4o, plus it does have an “o” in its name

openai.com/index/gpt-4-1/

This chart titled “GPT-4.1 family intelligence by latency” compares different variants of GPT-4.1 and GPT-4o models in terms of their:
• Intelligence (measured via Multilingual MMLU)
• Latency (speed of response)

Key Points:
• GPT-4.1 (top-right): Highest intelligence, highest latency.
• GPT-4.1 mini: A balanced option with solid intelligence and lower latency than full GPT-4.1.
• GPT-4.1 nano: Lowest intelligence but with the lowest latency—ideal for speed-critical tasks.
• GPT-4o mini: Positioned between 4.1 mini and 4.1 in intelligence, but with better latency than both—suggesting it’s a sweet spot for performance and speed.

The green diamonds mark GPT-4o mini variants, while blue circles indicate GPT-4.1 variants, emphasizing the efficiency gains of GPT-4o.

22

it was actually called GPT-4.10 but someone changed it to GPT-4.1

8

yes, GPT-4.5 does their marketing

3

More like this

1) What

my personal take on GPT-5 and LLMs in general is that we need to see a lot mo...

A medical paper from Microsoft lists the previously unknown model sizes of po...