Gemini 4 Neon ranks second on Arena.ai's Agent Arena Index behind Claude Fable 5.2
In brief
- Gemini 4 Neon scored 1336 on Arena.ai's Agent Arena Index across 4,200+ agent sessions.
- Claude Fable 5.2 (Max) was the only model ranked higher, scoring 1463.
- Neon's median cost per task on the index was $2.82, per Crypto Briefing.
- Neon API pricing, per Google: $3 per million input tokens and $15 per million output tokens.
Where Neon landed on the leaderboards
Neon also ranked second on the Agent Arena Index for steerability. Crypto Briefing defined that as how well a model adjusts when a user changes direction partway through a task (a pretty practical test if you're running agents on real work rather than on a static prompt).
Arena.ai's index wasn't the only leaderboard in the report. Neon took first place on Zapier's AutomationBench with a score of 56.3%, ranked second on the Vals Index and placed third on the Artificial Analysis Intelligence Index with a score of 61, according to Crypto Briefing.
The cost-per-task numbers
On the Agent Arena Index, Neon posted a median cost of $2.82 per task. Crypto Briefing said that works out to 53% cheaper per task than its nearest competitor, Claude Opus 5.5 (Max), and 69% cheaper than the top-ranked model on the index.
It's a median, and it's tied to this index.
Google's list pricing for the API is a separate figure. Crypto Briefing put it plainly:
Google lists Neon at $3 per million input tokens and $15 per million output tokens.
Fourth Gemini 4 release in seven weeks
Neon is the fourth release in the Gemini 4 series within seven weeks, the report said. That's a fast cadence. Earlier models in the family included Gemini 4 Carbon and Gemini 4 Hydrogen, and they've got their own marks on the same leaderboard: Carbon scored in the 1300s, while Hydrogen sat in the low 1200s.
Both of those sit below the 1463 that Claude Fable 5.2 (Max) posted on the Agent Arena Index. So does Neon's 1336, which is why it's second and not first on Arena.ai's overall ranking (and second again on steerability).
The rankings above are the ones Crypto Briefing reported from each benchmark's operator. We're not drawing a conclusion about which model is better.
Frequently asked questions
What does steerability mean on the Agent Arena Index?
Crypto Briefing defined steerability as how well a model adjusts when a user changes direction partway through a task. Gemini 4 Neon ranked second on the Agent Arena Index for steerability.
How much does Gemini 4 Neon cost to use?
Google lists Gemini 4 Neon's API pricing at $3 per million input tokens and $15 per million output tokens, according to Crypto Briefing. On Arena.ai's Agent Arena Index, Neon posted a median cost of $2.82 per task.


