Search

Google launches Gemini 4 Neon - faster, cheaper AI; African organisations need regional clarity

Google has released Gemini 4 Neon, a faster frontier model with a 1 million-token context window and token-based pricing.

A neon sign reading 'Gemini 4 Neon', an image credited to Google DeepMind used to announce the new model.
Google DeepMind

Google launches Gemini 4 Neon - faster, cheaper AI; African organisations need regional clarity

Published: 2026-11-18

Google has launched Gemini 4 Neon, a faster, lower-cost "frontier" model built for long, tool-heavy agent workflows such as software engineering, document synthesis and scientific analysis. Google DeepMind's blog reports Neon is priced at $3 per million input tokens ($0.15 per million cached input tokens) and $15 per million output tokens, and offers a 1 million-token context window.

Google's announcement says Neon is rolling out in the Gemini API in Google AI Studio, in Gemini for business (via the API and model dropdown), to paid Google AI Pro and Ultra subscribers in Google Antigravity, and to Pro and Ultra subscribers in the Gemini app. The company also publishes benchmark claims: Neon places #3 overall with a score of 61 on the Artificial Analysis Intelligence Index at xhigh, ranks #2 on the Vals Index and #2 on steerability in the Agent Arena Index, and is tested by Google to excel on multi-step agent workflows, coding, enterprise document synthesis and cybersecurity tasks.

Google Cloud's Agent Platform documentation lists Johannesburg (africa-south1) among the Agent Platform regions, and the documentation states that Agent Platform features vary by region and that not every feature is available everywhere. Organisations should confirm availability and feature support with their Google Cloud account teams.

That regional listing matters for African organisations because where inference and related platform features run can affect practical issues that matter to deployments. The Cloud pages highlight two relevant points:

  • Latency and user experience: Google Cloud documentation says processing close to end users reduces latency, so choosing a closer region can improve interactive or real-time performance.
  • Data residency and compliance: Google Cloud notes it stores and processes data in the region you specify for supported features, so which region handles inference affects where data is stored and processed and the compliance controls that apply.

Costs are a separate consideration: Neon is billed per input and output token at the published rates, and organisations should check with Google Cloud account teams for account-specific pricing and any regional or contract terms that affect overall costs for their workloads.

Google DeepMind's blog and Google Cloud location pages together are the primary sources for the model pricing and the region listing. The DeepMind blog reports the $0.15 per million cached input tokens figure; check those pages and your Google Cloud account team for the latest, account-specific pricing and feature availability.

African organisations and regulators should therefore check these specific points with cloud vendors or their Google Cloud account teams before deploying Neon in production:

  • Will Neon inference endpoints be available in the africa-south1 region for my account, and if so which features (model hosting, streaming, tool integrations) are supported there?
  • Where will inputs, cached inputs and outputs be stored and for how long? Are cached inputs guaranteed to remain in-region by contract?
  • What measured latency and availability trade-offs apply to running particular workloads from a region other than the users' location?
  • Do enterprise contracts include data residency, audit logs and breach notifications that meet local regulator requirements?

Gemini 4 Neon combines a very large context window with token pricing. Organisations should check region-level Neon endpoints and pricing details with Google Cloud account teams before deploying Neon in production. For the model announcement and pricing see Google DeepMind's blog: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/ and for regional feature availability see Google Cloud's Agent Platform locations: https://cloud.google.com/gemini-enterprise-agent-platform/machine-learning/general/locations.

Newzroom dot Africa has reached out to Irvine Partners and Information Regulator for comment and is awaiting a response. This story will be updated when they respond.

Source: Google DeepMind

Salute the work behind this story.

Blurbs

Up to 500 characters. Your name and location show with your blurb; your email is never shown. No links, email addresses or phone numbers in the blurb. Blurbs that break the terms of use are removed.

    More like this