Even the best AI models can't score above 60% on pure vision
Moonshot AI's new benchmark ranks GPT-5.6 Sol first at 59.7%, revealing that many "reasoning errors" are actually perception failures
00
Tag
vision-language-model
Matic's cleaning robot sees your home with 5 cameras instead of lidar
Tesla's 2019 vision-only bet has migrated to home robotics — Karpathy's argument, and lidar prices that have fallen to $200
00
Matic cleaning robot now goes where you point
An update combining voice and gesture matches arm direction to a 3D map to clean a specific spot
00
Pixel Watch 5 gains offline Gemini, price rises $50
Offline-capable Gemini and insulin resistance tracking added, price now $399
00
Google unveils SL2T, converts sign language to text, rolling out with Pixel 11
Instead of typing, deaf users can sign, and Gboard and Live Transcribe turn it into text
40
Cohere unveils 2.4B-parameter vision model 'North Micro Vision'
Ultra-compact VLM specialized in document and chart understanding, released free on Hugging Face under Apache 2.0
00
Liquid AI unveils screen-reading model that runs in 3GB
LFM2.5-VL-3B hits 20 tokens/sec on Galaxy S26 Ultra, 228 tokens/sec on M5 Max
00
LTX-2.5, open video generation model runs on a single consumer RTX GPU
An open-weight world model that produces a 10-second clip in 6.8 seconds, optimized for NVIDIA GPUs
70
Microsoft Research unveils CARE-X, which directly calculates cardiothoracic ratio from chest X-rays
Beyond generating reports, a method that measures heart and thoracic width to compute CTR
100
Microsoft Unveils Video-Capable Vision-Language Model Mage-VL
Multimodal model handling images, video, and text released open-source on Hugging Face
10
That's the last story.