Trending Now - Google's Gemma 4 Introduces Multi-token Prediction for Speedy AI

WHO:

Techmeme - Google is the organization involved in the release of the new technology.

MarkTechPost - Google AI is the specific division of Google responsible for the release.

WHAT:

Ars Technica - Google launched its Gemma 4 open models, promising a new level of power and performance for local AI.

Techmeme - Google released Multi-Token Prediction drafters for its Gemma 4 models.

MarkTechPost - The Multi-Token Prediction (MTP) drafters deliver up to 3x faster inference without quality loss.

WHEN:

Techmeme - Google launched its Gemma 4 open models in the spring.

HOW:

Techmeme - The MTP drafters use a form of speculative decoding to guess future tokens for faster inference.

MarkTechPost - Speculative decoding is the method employed to achieve the speedup in inference.

(UPDATED May 6, 2026 19:25 UTC)

Google has just launched its Gemma 4 open models, which promise to deliver a new level of power and performance for local AI. This development was reported at 19:25 UTC by Ars Technica. With these advancements, users can expect significant improvements in their AI capabilities.

--

(UPDATED May 6, 2026 17:32 UTC)

Google has just released Multi-Token Prediction drafters for its Gemma 4 models, which are designed to deliver up to three times faster inference without sacrificing quality. The new technology utilizes a method called speculative decoding to predict future tokens, significantly speeding up the processing time. Google AI is the specific division behind this innovative release, which was launched in the spring.