Brief IA
›

Google Launches Neon, Third in the Artificial Analysis Index

AI Tools·Tom Levy·

Google Launches Neon, Third in the Artificial Analysis Index

Google Launches Neon, Third in the Artificial Analysis Index
Key Takeaways
1Google is rolling out Gemini 4 Neon on its Gemini and Antigravity services.
2Artificial Analysis ranks it behind GPT-6.2 Astra and Claude Fable 5.2.
3A cyber variant is expected for eligible defenders in the coming days.
Why it matters : economic comparisons use Carbon's standard prices, while its promotion currently makes Neon 50% more expensive, according to Artificial Analysis.

Google deploys Gemini 4 Neon, ranked third by Artificial Analysis, in its Gemini API, its Gemini offering for businesses, Antigravity, and its Gemini application. The model sits between Hydrogen and Carbon.

Google Prepares Neon for Cyber Defenders

Eligible defenders are set to receive a cyber variant in the coming days through the Fairwind Program. This initiative from Google had already prioritized their access to Argon. On SEC-bench Pro, which evaluates defect detection in code, Google claims a score of 90.5%, with a cost per challenge roughly half that of its competitors.

The company assures that the protections inherited from Carbon against cybersecurity and biosafety abuses remain in place, with significantly fewer refusals to legitimate requests. For agents, Google also describes a coordination feature in Antigravity: Neon can simultaneously manage multiple Hydrogen sub-agents.

Performance and Pricing Compared to Carbon

Artificial Analysis estimates the current cost increase of Neon compared to Carbon at 50%. Until November 30, Carbon benefits from promotional prices of $2 and $10, while comparisons use its standard rates of $4 and $20. For Neon, Google charges $3 for one million incoming tokens, $15 for outgoing tokens, and $0.15 for cached entries.

However, Google presents a cost per task of $6.44 with Neon, compared to $7.76 with Carbon at standard rates. The company explains this discrepancy by the price of tokens, not by a reduction in responses: Neon has consumed about 15% more outgoing tokens per task. It also indicates a throughput of around 110 outgoing tokens per second for Neon, compared to about 57 for Carbon.

La lettre du soirGratuit · 19 h

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Google announces a score of 61 on the Artificial Analysis Intelligence Index with the very high reflection setting, one point ahead of Carbon. The independent ranking from Artificial Analysis places OpenAI's GPT-6.2 Astra, followed by Anthropic's Claude Fable 5.2, ahead of Neon. For programming agents, Google gives Neon a score of 59, compared to 58 each for Claude Fable 5.2 and Claude Opus 5.5. The corresponding costs per task are $6.86 for Neon, $12.61 for Opus, and $15.47 for Fable.

Neon Accessible to Pro and Ultra Subscribers

Google AI Pro and Ultra subscribers can choose Neon in Gemini as well as in Antigravity, Google's agent programming tool. In the Gemini application, Neon replaces Argon for the "In-Depth Analysis" option, according to available information. Current support remains with Hydrogen, while Carbon handles complex tasks.

Announced Wednesday by Koray Kavukcuoglu, Senior Vice President of Google DeepMind and Chief AI Architect at Google, Neon is the fourth model in Gemini 4 since the launch of Argon on September 30. Carbon followed on November 10, and Hydrogen two days later. Neon features a context of 1 million tokens and offers four reflection settings: low, medium, high, and very high. The high level is selected by default.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.