# Gemini 4 Neon at $3/$15: Vals and AA Back Google's Ranks, but Arena Now Has It Third

Mohit Tyagi · 2026-11-19 · Models

[https://timesof.ai/2026/11/google-gemini-4-neon-3-15-vals-second-aa-third-arena-third](https://timesof.ai/2026/11/google-gemini-4-neon-3-15-vals-second-aa-third-arena-third)

> Google released Gemini 4 Neon on November 18 at $3 per million input tokens and $15 per million output tokens, with a 1 million-token context window. It says Neon is third on the Artificial Analysis Intelligence Index, scoring 61 at xhigh, and second on the Vals Index. Vals lists it second of 42 models at 70.87%, and Artificial Analysis prints 61 for Neon at xhigh against 60 for Gemini 4 Carbon at high. Google also cites second place on Arena's Agent Arena Index at 1,336; Arena's board, read on November 19, shows Neon third at 1,333. Most other launch figures, including the cyber results, are Google's own.

Google's launch case for **Gemini 4 Neon** is that a mid-priced model now sits just behind the larger frontier models. The independent boards it cites mostly agree, though one rank had already slipped when we checked. Google released Neon on **November 18** at **$3 per million input tokens** and **$15 per million output tokens**, with a **1 million-token** context window.

Google says Neon is third on the Artificial Analysis Intelligence Index, scoring **61** at its xhigh setting, and second on the Vals Index. Vals lists it second of 42 models at **70.87%**. Artificial Analysis prints 61 for Neon at xhigh against 60 for [Gemini 4 Carbon](https://timesof.ai/2026/10/google-gemini-4-carbon-fairwind-cyber-frontier) at high. The trade is price: VentureBeat calculated that Neon costs 50% more per token than Claude Sonnet 5.5 or GPT-6.2 Sol and 25% less than Claude Opus 5.5, so the case for it depends on how many tokens a finished task takes.

## What Google released

Neon is rolling out in the Gemini API in Google AI Studio, in Gemini for business, to paid Google AI Pro and Ultra subscribers in Google Antigravity, and to Pro and Ultra subscribers in the Gemini app. Cached input costs **$0.15 per million tokens**. Google places Neon among its other Gemini 4 models: Carbon for the hardest long-horizon projects, [Hydrogen](https://timesof.ai/2026/11/google-gemini-4-hydrogen-everyday-agent-model-1-50-7-50) for fast everyday execution, and Neon, which it says "balances depth and efficiency." Carbon costs $2/$10 per million tokens until November 30 and $4/$20 afterwards, 9to5Google reported, which would make Neon 25% cheaper than Carbon from December 1.

Google says a cyber version of Neon will reach eligible Fairwind Program members over the next few days.

## Claims and checks

| Launch claim | Who measured | What the board shows |
| --- | --- | --- |
| Third on the Artificial Analysis Intelligence Index, 61 at xhigh | Artificial Analysis | Its page prints 61 for Neon at xhigh and 60 for Carbon at high. VentureBeat reports Artificial Analysis confirmed the third place on X; we did not read the full leaderboard. |
| Second on the Vals Index, 70.9% | Vals AI | Second of 42 at 70.87% (±0.94) with reasoning effort set to xhigh, and first on Vals's Finance Index (of 45) and Legal Index (of 43). |
| Second on Arena's Agent Arena Index at 1,336 over more than 4,200 sessions | Arena | The board, dated November 18 and read on November 19, lists Neon (High) third at 1,333, behind Claude Opus 5.5 (Max) at 1,336 and Claude Fable 5.2 (Max) at 1,448. |
| 32% lower cost per task than Carbon at the default setting | Google, citing Artificial Analysis | Artificial Analysis prints $3.74 per task for Neon at high and $7.76 for Carbon at high, a wider gap, but lists Carbon at $4/$20, not the $2/$10 introductory rate. Arena's median cost per task is $2.62 for Neon and $1.34 for Carbon, which it marks preliminary. |

On speed, Artificial Analysis lists 21.5 minutes per Intelligence Index task for Neon at xhigh against 37.2 minutes for Carbon at high, and an output speed of 110 tokens per second against 55. The settings are not matched, so the gap shows direction, not a like-for-like multiple. Vals says Neon sits on its score-versus-cost frontier on 7 of its 14 leaderboards.

A customer quoted in Google's announcement, Glean's Thai Tran, said Neon "matched or exceeded Gemini 4 Carbon on overall quality at 3x the speed and 60% of the cost" in early evaluations, VentureBeat reported, adding that Google's post gives no absolute completion times for a matched workload. At standard short-context rates, VentureBeat calculated, Neon would need roughly one-third fewer tokens than Sol or Sonnet to erase its 50% price premium.

## Limits

- The Terminal-Bench, FrontierCode, safety and cyber figures in the launch post are Google's own reports; we checked only Artificial Analysis, Vals and Arena.
- The boards use different effort settings, tasks and cost bases, so their ranks are not directly comparable. Arena's third place differs from the second Google cited, and we do not know whether Google's snapshot predates a reshuffle.
- We did not read Artificial Analysis's full leaderboard, so Neon's overall rank there rests on Google's claim and VentureBeat's report.
- The cyber version is not yet available outside the Fairwind Program, and we found no independent results for it.

## Sources

- [Google: Introducing Gemini 4 Neon (November 18, 2026)](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/)
- [VentureBeat: Google launches Gemini 4 Neon with Opus 5.5-level coding performance at 25% lower API token prices (November 18, 2026)](https://venturebeat.com/technology/google-launches-gemini-4-neon-with-opus-5-5-level-coding-performance-at-25-lower-api-token-prices)
- [9to5Google: Gemini 4 Neon launches: Carbon quality at Flash speeds, for AI Pro (November 18, 2026)](https://9to5google.com/2026/11/18/gemini-4-neon-announcement/)
- [Vals AI: Gemini 4 Neon benchmarks, cost and capabilities (November 18, 2026)](https://www.vals.ai/models/google_gemini-4-neon)
- [Artificial Analysis: Gemini 4 Neon vs Gemini 4 Carbon (read November 19, 2026)](https://artificialanalysis.ai/models/comparisons/gemini-4-neon-vs-gemini-4-carbon)
- [Arena: Agent Arena Leaderboard (snapshot dated November 18, 2026; read November 19, 2026)](https://arena.ai/leaderboard/agent)
