Bijan Bowen Tests Step 3.7 Flash Locally

Bijan Bowen30m
Bijan Bowen beside the headline LOCAL MODEL TEST on black.
0 comments · 0 votesOpen discussionClose discussion
Sign in to join the discussion

    Bijan Bowen introduces the model’s available formats and runs local tests on an M3 Ultra machineOn-device inference runs a trained model on the user's device rather than relying entirely on a remote model service., alongside a hosted comparison. He describes vendor claims separately from the tasks he actually tests.

    Bijan Bowen evaluates browser interfaces, a subway scene, games and an image-to-website task. Several outputs run with visible defects or require a follow-up fix, while other details impress him.

    Bijan Bowen observes that one four-bit result appeared better than selected eight-bit resultsModel quantization represents AI model values with fewer bits to reduce storage, memory use and often inference cost., but treats the small informal sample as inconclusiveEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria.. He also tests creative writing and notes mixed coding performance.

    Original YouTube thumbnailWatch on YouTube

    Video topic coverage

    • Agents & workflows0
    • Coding & tools85.6
    • Models & learning100
    • Infrastructure & compute78.4
    • Research & evaluation80.4
    • Safety & security0
    • Business & productivity0
    • Creative & media52.6
    • Robotics & perception0
    • Society & policy0

    Share this video

    Bijan Bowen beside the headline LOCAL MODEL TEST on black. With WWW.ARTIFICIAL-INTELLIGENCE.VIDEO, 1 June 2026 and duration 30m.

    Bijan Bowen explores Step 3.7 Flash with local and hosted configurations across coding and creative tasks.