Google is rolling out Gemini 4 Neon in the Gemini app today for AI Pro and AI Ultra subscribers. The model shows comparable or better performance to Carbon for code, reasoning, and agents, with speed comparable to Flash and lower cost.
Neon will be cheaper after the promotion
Neon pricing is $3/1M tokens received and $15/1M tokens generated. It represents twice that of Hydrogen. Until November 30, both Argon and Carbon remain billed less: $2/1M for entries, then $10/1M for exits.
Once this promotion ends, Argon and Carbon prices should increase to $4/1M for entries and $20/1M for exits. Neon will then be 25% cheaper than these models.
Subscribers see their selector change
In Gemini, Argon gives way to Neon, while version 3.1 is removed. The menu now combines “Deep Scan” at 4 Neon, “Daily Help” at 4 Hydrogen, and “Code & Complex Jobs” at 4 Carbon.
For paid accounts, 3.5 Flash-Lite appears under the heading “Auto”, presented by Google as a choice that adjusts to needs. Gemini tasks can also work with Neon. The model is also accessible from Google Antigravity to holders of a paid AI Pro or Ultra subscription.
Google aims Neon for complex professional work
After last week’s releases of Carbon and Hydrogen, Google places Neon between these models in Gemini 4. Hydrogen, Flash’s successor, remains aimed at quick daily activities; Carbon targets the most demanding projects over time, up to the limits of the model’s capabilities. Argon was previously offered for analysis and development, with Google recommending it for most software or knowledge work.
Neon is designed for complex development, analytical reasoning, and production business processes. Its context window reaches 1 million tokens. Like Hydrogen, it has four reflection settings: Low, Medium, High, and Very High.
The evening letterFree · 7 p.m.
Do you want the best AI tools before the rest?
We test and decipher the new AI tools every evening, in 5 minutes. Free.
Included upon registration: our selection of the best AI guides & comparisons.
Choose your pace
Free · No spam · Unsubscribe in 1 click
For writing, Google announces cleaner and better organized responses, with a direct and natural tone. The company also claims a more precise interpretation of imported tables, PDFs and photos, the transformation of extensive content into summaries or reports, as well as the extraction of relevant information from Gmail or Docs. According to Google, Neon regularly mobilizes the Web to update its responses, particularly when comparing prices between suppliers before a major purchase or preparing a summary incorporating recent market data.
Google also claims that complex, multi-step tasks run faster and more efficiently than larger models. The company promises more delegations before reaching usage caps. She cites the examination of invoices to explain an increase in costs, or the reconciliation of supporting documents and receipts to negotiate a reimbursement.
Neon scores 61 at Artificial Analysis
In the Artificial Analysis Intelligence Index, Neon comes in 3rd with 61 at the xhigh setting, ahead of comparable models but behind larger and slower models. Artificial Analysis also ranks it 1st in its Coding Agent Index, with 58.6. Neon obtains 65.2% on Terminal-Bench 4.1 and 46.2% on FrontierCode 2.0, the latter result corresponding to the high setting offered by default.
At Arena, Neon ranks 2nd in the Agent Arena Index worldwide: its score reaches 1336 over more than 4,200 real agent sessions. It is also in position 2 for the ability to take into account corrections made by the user during a task. On Agents’ Last Exam v1.1, he obtained 75.8%, a result at the level of the leading models.
Neon ranks 1st in Legal Research and Tax Agent ratings. On Surge AI, it comes first in most financial, accounting, table modeling and compliance with complex constraints tests. For Chartography, dedicated to dense professional graphics, he obtained 73.5%. Neon is also presented as the best in its category on Terminal-Bench Science with 70.5%, RiemannBench with 84.0% and HLE-Diamond with 71.3%.
In AutomationBench, Neon ranks 1st overall with 56.3%. He also finished 1st in Vending-Bench 2 from Andon Labs, a long-term business management simulation, with a result of $16,843.

