AINews news.smol.ai

PRIME: Process Reinforcement through Implicit Rewards

news.smol.ai·70w

not much happened today

news.smol.ai·70w

not much happened to end the year

news.smol.ai·71w

not much happened today

news.smol.ai·71w

not much happened today

news.smol.ai·71w

DeepSeek v3: 671B finegrained MoE trained for $5.5m USD of compute on 15T tokens

news.smol.ai·72w

not much happened today

news.smol.ai·72w

not much happened this weekend

news.smol.ai·72w

o3 solves AIME, GPQA, Codeforces, makes 11 years of progress in ARC-AGI and 25% in FrontierMath

news.smol.ai·73w

ModernBert: small new Retriever/Classifier workhorse, 8k context, 2T tokens,

news.smol.ai·73w

Genesis: Generative Physics Engine for Robotics (o1-mini version)

news.smol.ai·73w

Genesis: Generative Physics Engine for Robotics (o1-2024-12-17)

news.smol.ai·73w

OpenAI Voice Mode Can See Now - After Gemini Does

news.smol.ai·73w

o1 API, 4o/4o-mini in Realtime API + WebRTC, DPO Finetuning

news.smol.ai·73w

Meta Apollo - Video Understanding up to 1 hour, SOTA Open Weights

news.smol.ai·73w

Meta BLT: Tokenizer-free, Byte-level LLM

news.smol.ai·73w

Google wakes up: Gemini 2.0 et al

news.smol.ai·74w

ChatGPT Canvas GA

news.smol.ai·74w

OpenAI Sora Turbo and Sora.com

news.smol.ai·74w

Meta Llama 3.3: 405B/Nova Pro performance at 70B price

news.smol.ai·75w

Log in to enable infinite scrolling