Bijan Bowen Tests Gemma 4 12B for Local Coding

Bijan Bowen32m 2s
Bijan Bowen beside the words Gemma 4 12B.
0 comments · 0 votesOpen discussionClose discussion
Sign in to join the discussion

    Bijan Bowen introduces Gemma 4 12B and describes its local hardware requirements and encoder-free approach to audio and image input. He runs an 8-bit quantized model on a Mac StudioOn-device inference runs a trained model on the user's device rather than relying entirely on a remote model service. for the demonstration.

    His coding tests include a browser operating system, 3D printer simulation, visual website mockups and a C++ skate game. Several examples require corrections or manual help, while the game and interface experiments impress him.

    Bowen briefly tries audio input in the Mac application and uses another model to repair errors in a flight simulator. He presents the result as a promising first look rather than a controlled benchmarkEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria..

    Original YouTube thumbnailWatch on YouTube

    Video topic coverage

    • Agents & workflows0
    • Coding & tools98.9
    • Models & learning100
    • Infrastructure & compute81.3
    • Research & evaluation52.7
    • Safety & security0
    • Business & productivity0
    • Creative & media29.7
    • Robotics & perception0
    • Society & policy0

    Share this video

    Bijan Bowen beside the words Gemma 4 12B. With WWW.ARTIFICIAL-INTELLIGENCE.VIDEO, 3 June 2026 and duration 32m 2s.

    Bijan Bowen tests a local Gemma 4 12B model on coding and multimodal tasks.