GitSeonjeap
English

← News

Local AI

Perplexity Releases pplx-decider v1.3 and Flash v1.3 Models

Posted As of

Perplexity said on its Hugging Face model card and community announcement that it has released pplx-decider v1.3 and Flash v1.3. Accounts of the public access method and license differ between sources, so they need to be checked directly.

What happened

  • According to Perplexity's Hugging Face model card and community announcement (2026-11-12), v1.3 at 26B scale and Flash v1.3 were announced.
  • The card describes v1.3 as the successor to v1.2 and based on Qwen3.8-27B.
  • For self-hosting, the card says (official) that Python 3.12 or later and a CUDA GPU with about 49 GiB or more of memory for the BF16 weights are required.
  • The output is a probability rather than generated text, and v1.3 adds a locate function that extracts spans according to type or instruction.
  • Access method needs verification. The card text I checked says the repository is private and requires a token, but secondary reports describe public weights and an Apache 2.0 license, so the two do not match.

Our take

For people who want to run it locally on a Mac, PC, or glasses, the roughly 49 GiB memory requirement is a major burden. According to secondary reports, GGUF files are not yet available, so running it directly with llama.cpp or Ollama may be difficult. Public availability, license, and benchmark figures are still unconfirmed, and Ollama support for pplx-decider should be treated as a separate matter.

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