Subscribe

Applications

Where AI is actually being used: products, industries, and the work it changes.

  • MediaTek Dimensity 9500

    Two genuine firsts: hardware support for 1.58-bit ternary models, and a compute-in-memory NPU for always-on work.

  • Google Tensor G6

    A 50 per cent larger TPU on the same 3nm class, so the gain is architectural rather than a node shrink.

  • MediaTek Dimensity 9300

    MediaTek’s, the only explicit parameter ladder any mobile vendor has published, with named partner models.

  • Google Tensor G5

    The foundry switch to TSMC, and the first Tensor generation Google was willing to quantify properly.

  • Google Tensor G3

    The first Tensor architected for transformers, and the first to run Gemini Nano on device.

  • Google Tensor G4

    Memory capacity, not TPU throughput, was the enabling change: 16GB on Pro models let Gemini Nano stay resident.

  • Google Tensor

    Google’s first silicon, sized for its own speech and imaging models, and the least-disclosed generation of any chip on this site.

  • Google Tensor G2

    An incremental TPU uplift Google declined to quantify at all.

  • Snapdragon 8 Elite

    The naming break (no Gen number) and a step backwards in disclosure: the parameter ceiling published a year earlier disappeared.

  • Snapdragon 8 Elite Gen 5

    The real change is the precision set: INT2 and FP8 arrive, which matters more for modern quantised inference than the headline percentage.