
This is the reasoning model SmallThinker 3B running offline at 12 tok/s. Wild that in 2025 you can tell your phone to think.
Market Brief
Daily market recaps with key events, stock movements, and global influences
A newer version of this article is available. Read the latest version
9 posts • GPT (4o mini)
Updated
More breaking stories on DeepNewz — updated live.
This is the reasoning model SmallThinker 3B running offline at 12 tok/s. Wild that in 2025 you can tell your phone to think.
Google has caught on to RWKV-7 style models (my design with channel-wise factors is better)🤣Attention's days are numbered, because all we need for weak AGI is RNN + CoT + Tool Usage (e.g. external memory, external compiler, external verifier, web browser, ...)
Google AI Research Introduces Titans: A New Machine Learning Architecture with Attention and a Meta in-Context Memory that Learns How to Memorize at Test Time Google Researchers has proposed a novel neural long-term memory module designed to enhance attention mechanisms by…