Gavin Leech, UK-based AI researcher and co-founder of the consultancy Arb, notes vision models are roughly 1000 times smaller than text models owing to language's data compression through compositional semantics and abstractions

— Replies linked the gap to potential image-based chain-of-thought reasoning.

Original post

OPGL#1480gavin leech (Non-Reasoning)|@GLEECH

one of the major failures of my life was being so surprised to find out that vision models were ~1000x smaller than text models. Just total failure to understand language's god-tier data compression

2:00 AM · May 17, 2026 View on X

REPLYRO#1220rohit|@KRISHNANROHIT

@1a3orn CoT in pictures but not words would be quite neat

1A1a3orn@1a3orn

this is a relevant consideration for projecting how Lindy intelligible CoT is likely to be

2:31 PM · May 17, 2026 · 4.3K Views

4:15 PM · May 17, 2026 · 257 Views

QUOTE POST1A#13801a3orn|@1A3ORN

this is a relevant consideration for projecting how Lindy intelligible CoT is likely to be

GLgavin leech (Non-Reasoning)@gleech

one of the major failures of my life was being so surprised to find out that vision models were ~1000x smaller than text models. Just total failure to understand language's god-tier data compression

9:00 AM · May 17, 2026 · 189.2K Views

2:31 PM · May 17, 2026 · 4.3K Views