Trillion Labs (@TrillionLabs) / X

Trillion Labs

5 posts

Trillion Labs

@TrillionLabs

Superintelligence Lab based in Korea

Seoul, Korea

Joined January 2026

Trillion Labs reposted
Interconnects
@interconnectsai
Mar 3
Latest open artifacts (#19): @Alibaba_Qwen 3.5, @Zai_org GLM 5, @MiniMax_AI 2.5 — Chinese labs' latest push of the frontier. Featuring breakdown & analysis of: - Alibaba’s Qwen 3.5 (from 0.8B to 397B), Z.ai’s GLM-5 (744B), and @StepFun_ai 's Step-3.5-Flash.
8.5K
Trillion Labs
@TrillionLabs
Mar 11
Interconnects
@interconnectsai
Mar 3
Latest open artifacts (#19): @Alibaba_Qwen 3.5, @Zai_org GLM 5, @MiniMax_AI 2.5 — Chinese labs' latest push of the frontier. Featuring breakdown & analysis of: - Alibaba’s Qwen 3.5 (from 0.8B to 397B), Z.ai’s GLM-5 (744B), and @StepFun_ai 's Step-3.5-Flash.
huggingface.co
trillionlabs/Tri-21B-Think · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
249
Trillion Labs
@TrillionLabs
Feb 5
Replying to @TrillionLabs
Results across 6 benchmarks (MWMBench): → gWorld 8B: 74.9% avg accuracy (vs Llama 4 402B: 55.7%) → gWorld 32B: 79.6% → <1% render failure rate → Data scaling follows a power law (R² ≥ 0.94) Models, code, and benchmark all open: 🤗 8B: huggingface.co/trillionlabs/g… 🤗 32B:
103
Trillion Labs
@TrillionLabs
Feb 5
New World Model alert. This time it's mobile! Introducing gWorld — open-weight visual world models for mobile GUIs. Instead of generating pixels, gWorld predicts the next screen as executable web code. Render it → pixel-perfect image. gWorld 8B/32B beat frontier models up to
00:00
349