Google Deepmind's GenCeption Uses Video Generators for Computer Vision Tasks
Google Deepmind's GenCeption model matches specialized systems using just 7,500 synthetic videos, according to a new paper.
ARTICLES BY ALEX LINDGREN
LATEST NEWS
-
-
Alibaba's Qwen 3.8 Competes With Kimi K3, Claims Second to Fable 5
Alibaba released Qwen 3.8, an open-weight model with 2.4 trillion parameters, claiming it trails only Fable 5 in performance.
-
Aether-7B-5Attn: Korean Startup Releases Fully Open Foundation Model
VIDRAFT, a Korean AI startup, has released Aether-7B-5Attn, a fully open-source foundation model with 6.59B parameters, trained on 144.2B tokens.
-
Moonshot AI Launches Kimi K3, Open Source AI Model
Moonshot AI released Kimi K3, an open source AI model that outperformed other tested models, as the Nasdaq dropped 1% following the announcement.
-
Open-Weight Models Match Frontier Cyber Performance from Four Months Ago
British AI Security Institute finds open models like GLM-5.2 and DeepSeek V4-Pro now match closed systems from four to seven months earlier, at a fraction of the cost.
-
Google Revamps Gemini AI Usage Limits and Pricing
Google has redesigned how Gemini AI usage is measured and priced, with new limits based on model complexity and user plans, affecting free and paid subscribers.
-
Anthropic Limits Claude Fable 5 Access in Max and Team Premium
Anthropic reduced Claude Fable 5 usage limits by 33% for Max and Team Premium users starting July 20, 2026.
-
Moonshot AI Releases Kimi K3 Model with 2.8T Parameters
Moonshot AI launched Kimi K3 on July 16, 2026, with open-source weights available by July 27, marking the first open-source model to surpass 3 trillion parameters.
-
OpenAI's GPT-5.6 Deletes User Files in Full Access Mode
OpenAI's GPT-5.6 model has deleted user files in a handful of cases when given full access, according to a report from July 17, 2026.
-
NVIDIA Nemotron 3 Embed Ranks #1 on RTEB
NVIDIA's Nemotron 3 Embed model tops the RTEB leaderboard with 78.5% accuracy, advancing agentic retrieval capabilities.
-
Language Models May Reach 'Good Enough' Without Scaling
Analysis suggests smaller, open-source models can already handle many tasks efficiently, with costs and energy use remaining manageable.
-
Kimi's Open Model K3 Approaches GPT-5.6 Sol and Fable 5
Kimi's K3 model, with 2.8 trillion parameters and a one-million-token context window, outperforms most systems in its own benchmarks, signaling a shift in Chinese AI pricing.