AI News | 2 min read

Google Launches Gemini 4 Neon: Mid-Tier Model at $3 Per Million Tokens

Google released Gemini 4 Neon today, a production-grade mid-tier model priced at $3/million input tokens that replaces the older Argon and Gemini 3.1 Pro in its API lineup.

Hector Herrera
Hector Herrera
A Studio featuring document, related to an AI model 4 Neon: Mid-Tier Model at $3 Per Million Tokens
Why this matters Google released Gemini 4 Neon today, a production-grade mid-tier model priced at $3/million input tokens that replaces the older Argon and Gemini 3.1 Pro in its API lineup.

Google DeepMind released Gemini 4 Neon on November 19, 2026, expanding its Gemini 4 family with a new mid-tier model. The key detail: Neon ranks above Gemini 4 Carbon on the Artificial Analysis benchmark index while costing 32% less per task — a combination that reshapes the value proposition for the entire lineup.

Background

Google unveiled the Gemini 4 family earlier in 2026 with Carbon as its affordable tier and Hydrogen as its flagship. The gap between them left developers choosing between a fast, cheap model and an expensive, highly capable one. Neon fills that middle ground — but with benchmark scores that beat Carbon, the positioning reads more like a Carbon replacement than a true third tier.

The Gemini 4 models incorporate extended thinking — the ability to reason through problems step-by-step before responding, similar to what OpenAI introduced with its o-series models. Neon exposes four distinct thinking levels, letting developers tune how much compute to spend per query.

Specifications

  • Input pricing: $3 per million tokens
  • Context window: 1 million tokens — the same ceiling as Hydrogen
  • Thinking modes: Four levels, letting developers tune the latency-vs-depth tradeoff
  • Availability: Live in Google AI Studio and the Gemini API; rolling out to AI Pro and Ultra subscribers

Why the Cost Math Is Unusual

Neon being cheaper than Carbon while ranking above it on benchmarks is not typical. Usually, better performance costs more. The gap suggests one of two things: Google improved its training efficiency enough to deliver more capability at lower cost, or it priced Neon aggressively to compete against OpenAI's o-series mid-tier and Anthropic's Claude Sonnet family.

For developers processing large volumes, the 1M-token context at $3/M input is the headline number. A team summarizing 10,000 long legal documents per month — say, 100K tokens each — is processing one billion tokens. At $3/M, that's $3,000 per month on input alone. Comparable tasks on some competing mid-tier models run $4–5/M at similar context lengths.

The Gemini 4 Family at a Glance

Model Position Input Price Context Thinking
Hydrogen Flagship Not disclosed 1M tokens Yes
Neon Mid-tier $3/M tokens 1M tokens 4 levels
Carbon Entry Lower Undisclosed Limited

Google has not published full pricing for all tiers or Carbon's context window specifications, making a direct cost comparison difficult.

Impact

Businesses running document processing, long-form code generation, or multi-document research pipelines now have a third Google option that may undercut what they're paying elsewhere. Developers building on the Gemini API don't need to migrate — Neon is available immediately through the existing SDK.

Consumer subscribers to AI Pro and Ultra should see Neon become a default routing option for complex queries, potentially delivering faster or more accurate responses without additional cost. Google has not specified when the consumer rollout will complete.

What to Watch

Google has not released Carbon's full pricing or context window for a clean comparison. Watch for Carbon's positioning to shift — or for Google to consolidate the entry tier around Neon — as adoption data comes in over the next 60 days.

Key Takeaways

  • ✓ For developers processing large volumes

Did this help you understand AI better?

Your feedback helps us write more useful content.

Hector Herrera

Written by

Hector Herrera

Hector Herrera is an AI systems architect in Houston and founder of Hex AI Systems. He designs and runs AI systems in production and writes daily about how AI is reshaping business, government and everyday life. 20+ years building for the web. Houston, TX.

More from Hector →

Get tomorrow's AI briefing

Join readers who start their day with NexChron. Free, daily, no spam.

More from NexChron