←Story detail

Trending story

Survival LLM releases version 0.3.4

1 reports1 story updatesUpdated Nov 2 03:30

Understand this story

AI overview

Survival LLM releases version 0.3.4. According to Numerama, this is a free, open-source AI assistant that runs entirely offline, capable of answering survival questions such as first aid, water purification, and storing food during power outages, with sources cited for each answer. Released on October 27, 2026, this version automatically selects a model based on device memory: Macs with more than 128 GB of memory use the 80 GB NVIDIA Nemotron 3.5 Super, while machines with 16–31 GB of memory use the 6 GB IBM Granite 4.2 8B. In the developer's test of 22 safety and first-aid questions, the latter answered only 14 correctly, and each response consumed about 100 watts of power.

AI-generated from public reports

Latest development · Report

Survival LLM: a free survival first-aid AI assistant that runs entirely offline

Developments

1 confirmed developments

  1. Survival LLM releases version 0.3.4

    Source:Numerama

Reports and discussion timeline

Browse reports and discussion by time. Their evidence levels remain distinct.

Nov 2

  1. Numerama

    Survival LLM: a free survival first-aid AI assistant that runs entirely offline

    Survival LLM is a free, open-source AI assistant that runs entirely offline. Version 0.3.4 was released on October 27, 2026, and can answer survival questions about first aid, water purification, and storing food during power outages, citing sources for each answer. It automatically selects a model based on memory: Macs with more than 128 GB of memory use the 80 GB NVIDIA Nemotron 3.5 Super, while machines with 16–31 GB of memory use the 6 GB IBM Granite 4.2 8B. In the developer's test of 22 safety and first-aid questions, the latter answered only 14 correctly, and each answer consumed about 100 watts of power.

Related stories

  1. Related storiesSurvival LLM releases offline AI emergency assistant
  2. Related storiesDeepSeek releases DeepSeek-V4.1-Flash-1017
  3. Related storiesEnglish translation is being prepared.
  4. Related storiesMiniMax announces that M3.1 Flash is supported by multiple community quantization solutions for local operation