Google Tensor G3 chip powers AI upgrades in Pixel 8 phones

The Pixel 8 and Pixel 8 Pro launch with Google’s third‑generation Tensor G3 system‑on‑chip, which Google says can run more than twice the on‑device machine‑learning models as the Pixel 6 and handle models ten times more complex [1][4]. The upgrade is positioned as the hardware foundation for a wave of AI‑driven features, from enhanced Call Screen to on‑device generative AI.

AI performance gains

Google claims the Tensor G3 enables the Pixel 8 series to run over twice as many on‑device machine‑learning models as the Pixel 6, and the largest model is ten times more complex [1][4]. This translates into new capabilities such as faster, more accurate Call Screen spam detection, higher‑quality audio and video via Clear Calling, and the “Audio Magic Eraser” that removes unwanted sounds from recordings. The chip also powers a new screen‑reading shortcut and faster playback controls, all driven by the upgraded TPU and ISP built into the SoC.

Hardware details and efficiency

A leaked benchmark shows the G3 is built around a 9‑core CPU with a Cortex‑X3 core clocked at 2.91 GHz, the same high‑end ARM core found in Snapdragon 8 Gen 2, and paired with 12 GB of RAM and a Mali‑G715 GPU [1]. The chip is fabricated on Samsung’s 4 nm process, promising better power efficiency than the previous Tensor G2. Google’s own blog notes that the new TPU is custom‑designed for on‑device generative AI, and the overall architecture is co‑designed with Google Research to handle models 150 times more complex than those on the Pixel 7 [4].

Competitive context

While the Tensor G3 delivers notable AI improvements, analysts have long noted that Google’s custom silicon lags behind Qualcomm’s Snapdragon and Apple’s A‑series in raw performance. The upcoming Tensor G6 and G7 chips are rumored to move to 2 nm and 2 nm processes respectively, but early leaks suggest only modest gains over earlier generations [2][3]. This underscores Google’s strategy of compensating for hardware limits with software‑driven AI features.

What to watch

• Feature rollout – Monitor the activation of on‑device generative AI tools (e.g., Magic Eraser, Clear Calling) in the weeks after launch.
• Benchmark updates – Look for independent performance tests comparing Tensor G3 against Snapdragon 8 Gen 2 and Apple A‑series chips.
• Future chip roadmap – Leaks of Tensor G6 and G7 (targeting 2 nm) could signal whether Google can close the performance gap in upcoming Pixel models [2][3].

The Tensor G3 marks Google’s most aggressive push to embed advanced AI directly in its flagship phones, but the longer‑term impact will hinge on whether software advances can offset the chip’s relative performance lag behind rivals.