Back to News

Hugo Vergnes trains 3.8B LLM to 0.384 CORE for $998

118 points · 21 comments#llm#training#efficiency#cost

Hugo Vergnes reports training a 3.8-billion-parameter language model to a 0.384 CORE benchmark score for $998. The post details the cost breakdown and the techniques used to achieve this efficiency.

Coverage timeline

  1. Hacker NewsAnon84
    Training a 3.8B LLM to 0.384 CORE for $998 – Hugo VergnesHN

Related stories

  • Average LLM Token Cost Drops to 97 Cents, Down From May High of $2.07

    A closely watched index of AI token prices fell to a fresh low this week, with the average cost per million tokens now at 97 cents. This continues a sharp months-long decline from a peak of $2.07 in late May, reflecting deflating prices in the AI sector.

    1 reports · 1 sources

  • Tiny LLM Hits 21,000 Tokens/s on $250 FPGA in Live Demo

    A Hacker News post showcases a live demo of a tiny language model running at 21,000 tokens per second on a $250 FPGA. The demonstration highlights the potential for low-cost, high-speed inference on edge hardware.

    1 reports · 1 sources79 points · 30 comments

  • Z.ai debuts GLM-5.3 with post-training scaling gains, open weights in two weeks

    Z.ai (Zhipu AI) announced GLM-5.3, an open-source model built on the same base as GLM-5.2 with substantially extended post-training, yielding a 50% gain on internal coding benchmarks and state-of-the-art results among open models on Terminal Bench 3.0 and Agents' Last Exam (CLI). The model matches Mythos 5 in white-box code review and vulnerability discovery, and scored 60 on the Artificial Analysis Intelligence Index, tying Kimi K3. Model weights will be released in two weeks after security hardening; API pricing is set at $1.40 per million input tokens and $4.40 per million output tokens.

    9 reports · 6 sources1154 points · 573 comments

News updates every hour. Subscribe in your RSS reader to get new stories automatically.

Subscribe