Gemini 4 Neon arrives with Carbon-level performance at lower cost

Google released Gemini 4 Neon on November 18, 2026, adding a model to a Gemini 4 line that already included Argon, Carbon and Hydrogen. According to Google's launch post, Neon sits between Carbon, which targets long and demanding projects, and Hydrogen, which handles fast everyday work. Google says Neon ranks third overall on the Artificial Analysis Intelligence Index with a score of 61 at its highest thinking setting, and second on the Vals Index. The model offers four thinking levels and a 1 million-token context window.
Access is split by product. AI Pro and Ultra subscribers can select Neon in the Gemini app, where it takes Argon's place in the model picker, and Gemini 3.1 Pro is being retired. Developers can reach it through the Gemini API and Google AI Studio, while Gemini for business and Antigravity also carry it. List pricing is $3 per million input tokens and $15 per million output tokens, with cached input at $0.15, which is double the $1.50 and $7.50 charged for Hydrogen. Carbon and Argon are discounted to $2 and $10 until November 30, then rise to $4 and $20, according to ITdaily, so Neon is the most expensive Gemini 4 model per token until then and roughly 25% cheaper than both from December 1.
Why it matters
For companies building on Gemini, the choice now centers on which tier to make the default. Neon costs twice as much per token as Hydrogen, so teams that route routine work to Hydrogen would see higher unit costs if they moved that work up. Trending Topics reports an average Neon index task at $6.44, compared with $7.76 for Carbon at its high setting, and lists coding-task averages of $6.86 for Neon against $12.61 for Opus 5.5 and $15.47 for Fable 5.2. Those figures are third-party and depend on test conditions.
For investors and operators, cost per completed task matters more than the list rate. Thinking tokens are billed at the output rate, so higher thinking settings raise spend even when the price per token stays flat. The end of Carbon's promotion on November 30 and the retirement of Gemini 3.1 Pro both create migration work and new evaluation budgets. Google's lower fabrication rate comes from its own internal testing, so teams should measure it on their own data before changing defaults.
Key points
- 1Neon launched November 18, 2026 in the Gemini API, Google AI Studio, Gemini for business and Antigravity.
- 2Neon is priced at $3 per million input tokens and $15 per million output tokens, with cached input at $0.15.
- 3Carbon and Argon rise from $2 and $10 to $4 and $20 per million tokens on December 1.
- 4Google reports a score of 61 at the highest thinking setting on the Artificial Analysis Intelligence Index, ranking third overall.