Excels on SWE-bench (incl. multilingual), Terminal-bench, and long-horizon tasks. Try it on Poe across web, desktop, mobile — and via the Poe API at https://
poe.com/MiniMax-M2.1 (2/2)
INNOVATION
-
MiniMax-M2.1 Excels on SWE-bench and Long-horizon Tasks
By
–
-

MiniMax M2.1: Agent-Optimized AI Model for Coding and Planning
By
–
Now on Poe: MiniMax M2.1 An agent-optimized model built for coding, tool use, and long-horizon planning served via @novita_labs (1/2)
-
Future Programmers Will View Today’s Tech Experts as Outdated
By
–
The undergrads of today will look at us like we looked at VHDL experts
-
Robotic Steering: Mechanistic Interpretability for VLA Finetuning
By
–
🎉Despite massive pretraining, VLAs need to adapt to specific physical contexts. We introduce Robotic Steering, a novel finetuning method using mechanistic interpretability to surpass standard FT:
— Chancharik Mitra (@chancharikm) 26 décembre 2025
🎁 22× fewer parameters
🎁 +53% on unseen tasks
🎁 Interpretable
Thread below👇 pic.twitter.com/RsSU8RANNq🎉Despite massive pretraining, VLAs need to adapt to specific physical contexts. We introduce Robotic Steering, a novel finetuning method using mechanistic interpretability to surpass standard FT: 🎁 22× fewer parameters 🎁 +53% on unseen tasks 🎁 Interpretable Thread below👇
-

OpenAI launches Sora 2 with improved realism and native audio
By
–
OpenAI: Sora is in broad rollout, and Sora 2 is live
OpenAI’s Sora has been publicly accessible via http://
sora.com, and Sora 2 adds stronger realism and native dialogue/sound effects. OpenAI also calls out watermarking + provenance signals (including C2PA metadata) -
AI Tools Move From Demos to Production Workflows
By
–
What this means: we’re moving from “cool demos” to production defaults.
The winning teams in 2026 will be the ones who standardize these tools into repeatable workflows: templates, brand consistency, and faster iteration loops. -

Stability AI Advances Text-to-Image Generation with SD 3.5
By
–
Stability AI: text-to-image keeps leveling up
Stability’s SD 3.5 lineup and Stable Image models focus on higher-fidelity outputs, better prompt adherence, and options tuned for speed vs quality (including “Ultra” and faster variants). https://
buff.ly/nqNrGwy -

Google Veo 3.1 Refresh: Enhanced Video Generation with New Features
By
–
Google: Veo 3 got refreshed (Veo 3.1)
Google shipped Veo 3.1 + 3.1 Fast with upgrades like: more options for clip durations (4/6/8 seconds) image referencing (up to 3 images) and first/last frame controls “extend” style workflows for Veo-created videos -

Creative AI Gets Major Upgrade Cycle for Content Creators
By
–
Heads up: creative AI just got a real upgrade cycle. If you make videos, ads, socials, or design assets. These 5 updates are worth a quick scan.
-

Canva AI Animation Speeds Up Polished Video Creation
By
–
Canva: AI animation + faster video creation
Canva’s leaning harder into “make it move” workflows, with AI-assisted animation choices and built-in AI video generation so non-editors can ship polished clips faster. https://
buff.ly/aPdndbw