THE CRUNCH
Google has released Gemini 4 Neon, which the company pitches as delivering "frontier-level intelligence" for demanding software engineering, knowledge and science work. The model is live through the API at $3 per million input tokens and $15 per million output tokens. Benchmarks have now arrived: at its highest reasoning level, "xhigh", Neon scores 61 on Artificial Analysis' Intelligence Index, edging past Gemini 4 Carbon at 60 and landing third behind OpenAI's GPT-6.2 Astra and Anthropic's Claude Fable 5.2, both of which Google omits from its own comparison table.
Neon is also notably quicker. It generates around 110 tokens per second, nearly double Carbon's rate of 57, cutting benchmark processing time from about 36 minutes to roughly 21 minutes, according to Artificial Analysis figures. The model takes in text, images, video, audio and PDFs, with a context window of roughly one million tokens and up to a million output tokens. It can drive a computer, run code, search files and ground answers using Google Search and Maps, though it cannot generate images or audio.
One framing needs correcting from earlier coverage: The Decoder now describes Neon as Google's fourth AI model in seven weeks, not four. The Register dates the previous Argon release to late September, with Carbon following two weeks before Hydrogen opened to developers last week. Neon's $3/$15 pricing sits 25 percent below Carbon's regular rate of $4/$20, which takes effect on 1 December, but 50 percent above the promotional rate still in place through the end of November.
Google's positioning of the family remains loose. DeepMind's Koray Kavukcuoglu says Neon "balances depth and efficiency" between Carbon's long-horizon projects and Hydrogen's fast everyday execution. As The Register observes, the element naming scheme, unlike OpenAI's celestial ladder, does not signal a hierarchy, and Google's Antigravity tooling can swap models on the fly, delegating billing-affecting routing decisions to the service.
Side by side
Neon vs Carbon
| Neon | Carbon | Difference | |
|---|---|---|---|
| Intelligence Index | 61 | 60 | – |
| Input price | $3 per million tokens | $4 per million tokens | +33% |
| Output price | $15 per million tokens | $20 per million tokens | +33% |
Details from The Decoder, The Register, as quoted in this brief.


