Happy to say I was quite wrong
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex
I predict 48 for GLM 5.2
7:20 AM · Jun 17, 2026 · 14.9K Views
AI commentator Teortaxes concedes incorrect prediction as GLM-5 scores 40 on the Artificial Analysis index
Story Overview
AI commentator Teortaxes publicly conceded an overly optimistic forecast after GLM-5 posted a 40 on the Artificial Analysis Intelligence Index, missing the 48 he had projected for the GLM 5.2 variant.
821311215.2K
Model Iteration
Later variants moved past the miss
GLM-5.2 has since reached 51 on the same index, shifting attention from the earlier shortfall to steady gains in the GLM series for open-weights agentic work.
Open Question
Predictions remain a moving target
Public calls on new model scores continue to overshoot or undershoot live benchmark updates, leaving room for revised expectations as Zhipu iterates.
Sentiment
Users criticized xAI models for failing basic tests like text art and questioned data team priorities after an analyst admitted overestimating GLM-5 scores, while one expressed happiness for techshrek.
Pos
25.0%
Neg
75.0%
4 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Posts from X
Most Activity
Most Activity
VIEWS1.1KLIKES10
kache@yacineMTB
@teortaxesTex I am so happy for techshrek
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex
Happy to say I was quite wrong
2hViews 1.1KLikes 10Bookmarks 0
REPLIES1
Satyam Kumar ☄️@STUD_MAN_X
@teortaxesTex closest guess
1hViews 42Likes 1
testtm@test_tm7873
@teortaxesTex @STUD_MAN_X 😎
1hViews 18Likes 2
SimpsonsDNA@W4bbitSeason
@teortaxesTex I expect you to make a good Zigger joke about this.
2hViews 135
avturchin - e/resurrect@turchin
@teortaxesTex Failed my text art test as it can't even output any text art.
2hViews 80
Satyam Kumar ☄️@STUD_MAN_X
@teortaxesTex 😭 next - qwen 3.8 max
1hViews 72
RaphV@RaphV9
@teortaxesTex For me the most relevant benchmark is Terminal-Bench Hard
16mViews 48
Lunexa@Lunexalith
@teortaxesTex Wtf is wrong with http://X.ai ? More than 100,000 GB200 equivalent cards!
2hViews 46
Janek Mann@janekm
@teortaxesTex I wish they had also reported for "High" not just "Max", that might have edged it closer to the green quadrant.
1hViews 17
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex
@STUD_MAN_X not as close as this guy
1hViews 12
caio temer@canalCCore2
@teortaxesTex quando eu vi, pensei nos seus 48 na hora kkkk
2hViews 11
Mysterious Bishounen Gun Fighter Rider@666Decadence666
@Lunexalith @teortaxesTex It's what happens when your data team is focused on making the dataset less woke
51mViews 2Likes 1
Janek Mann@janekm
@turchin @teortaxesTex Maybe that's the kind of task for a VLM will do better.
1hViews 3