All tags

Topic: "tokenization"

    Gemini 2.0 Flash GA, with new Flash Lite, 2.0 Pro, and Flash Thinking
    Meta BLT: Tokenizer-free, Byte-level LLM
    Llama 3.2: On-device 1B/3B, and Multimodal 11B/90B (with AI2 Molmo kicker)
    Grok 2! and ChatGPT-4o-latest confuses everybody
    Chameleon: Meta's (unreleased) GPT4o-like Omnimodal Model
    Google I/O in 60 seconds
    GPT-4o: the new SOTA-EVERYTHING Frontier model (GPT4O version)
    Llama-3-70b is GPT-4-level Open Model
    Meta Llama 3 (8B, 70B)
    DBRX: Best open model (just not most efficient)
    Mistral Large disappoints
    Karpathy emerges from stealth?
    The Core Skills of AI Engineering
    1/13-14/2024: Don't sleep on #prompt-engineering
    12/28/2023: Smol Talk updates