@teortaxesTex Speciale does use like 20k more tokens on average
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex
@scaling01 V3.2 Speciale still being #10 is very funny
1:27 PM · Jun 27, 2026 · 768 Views
LisanBench creator Lisan al Gaib says DeepSeek V3.2 Speciale consumes 20,000 more output tokens on average than peer reasoning models
The high token consumption leaves the model tenth overall.
2800900
Sentiment
Sentiment building, check back later.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Posts from X
Most Activity
Most Activity
VIEWS167LIKES1
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex
@scaling01 valid
Lisan al Gaib@scaling01
@teortaxesTex Speciale does use like 20k more tokens on average
3hViews 167Likes 1Bookmarks 0
Lisan al Gaib@scaling01
@teortaxesTex I just realized Kimi K2 is still cooked and has 0 tokens
whatever
3hViews 57Likes 1