Artificial Analysis: Gemini 4 Neon ranks third on Intelligence Index with 61 points
In independent testing by Artificial Analysis, Google’s new Gemini 4 Neon at xhigh scored one point more than Carbon at high. Neon is about 17 per cent cheaper per task, but only against Carbon’s standard prices.
Google released Gemini 4 Neon on 18 November. The company places it in its model family between the fast Hydrogen and Carbon, which is meant for the most demanding work.
The Artificial Analysis Intelligence Index combines nine benchmarks covering areas such as agentic knowledge work, coding and scientific reasoning, according to an image in the post. At its highest reasoning effort, xhigh, Neon scored 61 points. Reasoning effort sets how much the model thinks before answering. GPT-6.2 Astra and Claude Fable 5.2 rank ahead of Neon.
The table compares Neon with Carbon using figures from Artificial Analysis’s post on X.
| Metric | Neon (xhigh) | Carbon (high) |
|---|---|---|
| Intelligence Index | 61 | 60 |
| Cost per task | $6.44 | $7.76 |
| Time per task | approx. 21 min | approx. 36 min |
| Output speed | approx. 110 tokens/s | approx. 57 tokens/s |
| Price per 1M tokens (input/output) | $3 / $15 | $4 / $20 |
Source: Artificial Analysis, 18 November 2026. Carbon’s cost is calculated at standard prices. Times and speeds are approximate.
A one-point gap does not show that Neon is clearly better than Carbon. Nor does the lower cost come from efficiency: Neon produces about 15 per cent more output tokens per task. The saving comes entirely from token prices that are 25 per cent lower, Artificial Analysis said. Carbon’s discounted price of $2 / $10 applies until 30 November. Until then, Neon’s tokens cost 50 per cent more than Carbon’s.
Google’s announcement says Neon’s cost per task at the default high reasoning effort is 32 per cent lower than Carbon’s. Artificial Analysis said that at high, Neon matches Carbon for less than half of Carbon’s cost at standard prices. Google does not say what its comparison is based on.
In coding, results also depend on the harness: the program that steers the agent, gives the model its tools and runs its steps. On Artificial Analysis’s coding agent index, Neon (xhigh) scored 59 points in Google’s Antigravity CLI. Claude Fable 5.2 (xhigh) and Claude Opus 5.5 (max) scored 58 points in Claude Code. An average task cost $6.86 with Neon, $12.61 with Opus and $15.47 with Fable. Google said Neon scored 12.8 per cent on SWE-sweep in the mini-SWE harness and 39.0 per cent in Antigravity.
Cached input tokens cost $0.15 per million. The context window holds 1 million tokens. The model is available in Gemini for business and on the paid tier of the Gemini API in Google AI Studio. Google AI Pro and Ultra subscribers can also use it in the Gemini app and Antigravity.
Read more: Models