GitSeonjeap
English

← News

Local AI

Edge0-145B Open Weights Released, Claims On-Device Execution

Posted As of

Edge0 announced the release of open weights for Edge0-145B, a 145B-parameter model. The figures on the official model card are based on the RTX 5090 D, and macOS support differs across sources, so it needs to be verified.

What happened

  • According to the official Edge0 blog, Edge0-145B was introduced as an open-source on-device AI model with 145B parameters.
  • (Official) NVFP4 model card figures: agent decode 98.3 tok/s, 16K prefill 3165.7 tok/s, Terminal-Bench 2.1 score 86.5. The hardware basis for these figures is an RTX 5090 D 24GB.
  • According to the model card, the architecture has 48 layers, 512 experts, 10 active per token, a 256K context, and native 4-bit training.
  • For commercial use, the page says to contact info@edge0.ai.
  • According to the written content, Qwen's separate approval and attribution requirements are not waived by Edge0's permission.

Our take

For users who want to run models locally on a Mac or PC, the key question is whether a model at the 145B scale can run on-device. However, macOS support is inconsistent between secondary reports and the model card description, so it cannot be confirmed yet. Memory usage and speed figures for laptops were not confirmed in the original blog post, and the commercial use terms also require an inquiry.

Sources and related posts

More in Local AI

  1. NVIDIA Releases Nemotron for Science for Scientific Agents (nvidianews.nvidia.com)

    NVIDIA announced a package based on Nemotron 3.5 Super VL. It includes a chemistry dataset and a post-training recipe, and the IR copy states that the model is optimized to run locally on a 128GB DGX Spark.

    Local AI| |1 sources |Our take included
  2. NVIDIA RTX Spark Compact Desktops Begin Shipping (blogs.nvidia.com)

    Compact RTX Spark desktops from ASUS, Dell, and MSI have launched as of the announcement date and are positioned for local AI agents and model fine-tuning. Pricing was not confirmed on NVIDIA's page.

    Local AI| |1 sources |Our take included
  3. llama.cpp b12497 adds tiny-voca support (github.com)

    llama.cpp pre-release b12497 adds tiny-voca support to the multimodal path and improves G2P conversion. The release page has no performance figures or description of the target model.

    Local AI| |1 sources |Our take included

All news