Google launches Gemini 4 Neon for faster coding and research
Google says Gemini 4 Neon brings a million-token context window, lower cached-input pricing and faster performance to coding and research workflows.
Google has announced Gemini 4 Neon, a new model aimed at complex software engineering, knowledge work and scientific workflows. Google says Neon is designed to complete long, multi-step tasks quickly, rather than simply answer short prompts. The model is available from 18 November through the Gemini API in Google AI Studio and Gemini for business, while paid Google AI Pro and Ultra subscribers can use it in Google Antigravity. Pro and Ultra subscribers can also access it in the Gemini app, according to Google.1
The headline specification is a 1-million-token context window. That gives Neon room to work across large collections of documents, spreadsheets, charts, audio and video in one task. Google says the model can turn those mixed inputs into structured deliverables, which is the company’s pitch for enterprise and research use rather than a promise of independent performance in every workflow.1
Pricing is set at $3 per million input tokens, with cached input priced at $0.15 per million tokens, and $15 per million output tokens. Google says the cached rate represents a 95% discount. UK pricing, regional availability and the limits attached to paid consumer plans were not detailed in the announcement, so British users will need to check the relevant Google account or API billing page before treating the US-dollar figures as a final local cost.1
Google’s announcement also highlights benchmark results. It says Neon ranks third overall on the Artificial Analysis Intelligence Index at its xhigh thinking level, second on the Vals Index and second on the Agent Arena Index. Those are reported benchmark positions rather than independent testing by iHeartGeek. The company also says Neon matched or exceeded Gemini 4 Carbon in internal evaluations while running three times faster and costing 60 per cent less; that comparison is likewise a Google-reported result.1
For developers, Google is positioning Neon around production code and agent workflows. The company says it can handle large codebases, tool-heavy sessions and parallel work in Google Antigravity, with an emphasis on following corrections during a task and reducing duplicated file reads. It also reports strong results across coding, research, science and mathematical evaluations, although the selection and conditions of those tests vary between benchmarks.1
Availability and limits
The launch is broad, but it is not a guarantee that every feature is available everywhere. Google names AI Studio, Gemini for business, Antigravity and the Gemini app as launch destinations, with different account requirements for each. The announcement does not provide a UK-specific rollout timetable or consumer-plan allowance, and it does not establish how Neon compares with competing models under a common independent test set.1
Our opinion
Gemini 4 Neon is interesting because Google is selling speed, context and cost together instead of treating a larger model as the only answer. A million-token window and lower cached-input pricing could matter to teams processing the same large project repeatedly. But the most striking performance claims remain the company’s own launch figures, while real-world API bills will depend on how much context applications send and how many outputs they request. The sensible early view is that Neon looks like a serious new option for long-running technical work, not proof that Google has settled the wider AI race.
Google’s official announcement is the primary source for the model’s availability, pricing and benchmark claims.1