# Google prices Gemini 4 Neon at $3 and $15 per million tokens, 25% below what Gemini 4 Carbon will cost from December 1

Early third-party tests put Neon one point above Carbon on the Artificial Analysis Intelligence Index, at about half of Carbon's cost per task.

- Published: 2026-11-20T06:11:13.056Z
- Canonical: https://polylog.news/ai/2026-11-20/google-prices-gemini-4-neon-at-3-and-15-per-million-tokens-2
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Google DeepMind](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/)

Google DeepMind [released Gemini 4 Neon](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/) on November 18. Coverage of the launch agrees on [the price](https://www.mindstudio.ai/blog/gemini-4-neon-pricing-benchmarks): $3 per million input tokens, $15 per million output tokens and $0.15 for cached input, with a context window of about one million tokens and no free API tier. Neon is also included in Google's consumer AI Pro and Ultra subscriptions.

The price matters because of Google's own lineup. Gemini 4 Carbon and Argon are discounted to $2 and $10 per million tokens through November 30, and Carbon then rises to $4 and $20, [according to launch coverage](https://explainx.ai/blog/gemini-4-neon-launch-price-benchmarks-vs-carbon-hydrogen-november-2026). From December, Neon will therefore cost 25% less than Carbon, but for the next ten days it costs more than Carbon's discounted rate.

Independent figures are early and do not fully agree. [Neurohive reports](https://neurohive.io/en/news/gemini-4-neon-benchmarks-pricing-carbon-comparison/) that Neon scores 61 on the Artificial Analysis Intelligence Index, one point above Carbon, while another listing gives Neon 59.6 at its high setting. Artificial Analysis measured Neon at $3.74 per task at the high setting and $6.44 at the extra-high setting, compared with $7.76 for Carbon at high. Single sources report 65.2% on Terminal-Bench 4.1, a top score of 58.6 on the Artificial Analysis Coding Agent Index, and 56.3% on Zapier's AutomationBench. Treat these numbers as provisional until the Artificial Analysis leaderboard settles.

Other vendors price in different bands. Alibaba launched Qwen4-Max on November 16 as an API-only model at [$1.50 and $4.50 per million tokens](https://www.datacamp.com/blog/qwen4-max), with no downloadable weights, and it scores [43 on the same Artificial Analysis index](https://www.siliconreport.com/qwen4-max-scores-43-on-artificial-analysis-intelligence-index). At the top end, Anthropic's gated Claude Mythos 5.2 lists at a [reported $8 and $40](https://www.eesel.ai/blog/claude-mythos-5-2).

## What this means

The common price-war framing misses the mechanism here. Google is not cutting the price of its top tier. It is raising Carbon's price on December 1 and placing a near-equal model below it, so customers who do not migrate will pay more. Teams building agents with high token volume gain if the near-parity with Carbon holds on their own workloads, and they lose if it holds only on aggregate indexes. Google can sustain this because it serves Gemini on its own tensor processing units (TPUs), while API resellers and labs that rent compute have less margin available. Alibaba's decision to keep Qwen4-Max closed cuts against the desk's tracked thesis that Chinese open weights are becoming the non-US default at the frontier. The likelier outcome within weeks is that OpenAI and Anthropic answer with a cheaper tier or a temporary discount, not a cut to their flagship prices. This depends on Neon's lead over Carbon surviving independent runs. If Artificial Analysis settles Neon below Carbon, the move becomes a price increase presented as a new tier, and the competitive pressure on rivals goes away.

## What to watch

- Neon's final placement on the Artificial Analysis leaderboard after the launch-day numbers settle. A finish below Carbon would turn the release into a reshuffle of Google's tiers, not a cost advantage.
- Whether Google extends Carbon's discount past November 30. An extension would show weaker demand for Neon than Google expected.
- Whether Alibaba follows through on open weights for smaller Qwen4 models. A further delay would mean Chinese labs are holding back their best models, as US labs do.
