Gemini 4 Neon takes second place on Agent Arena Index, trailing only Claude Fable 5.2 Google DeepMind's Gemini 4 Neon placed second overall on Arena.ai's new Agent Arena Index with a score of 1336, drawn from more than 4,200 real-world agent sessions, trailing only Anthropic's Claude Fable 5.2 (Max) at 1463. Launched November 18, 2026, Neon posted a median cost of $2.82 per task on the index — 53% cheaper than Claude Opus 5.5 (Max) and 69% cheaper than the top-ranked model — with Google listing pricing at $3 per million input tokens and $15 per million output tokens. Neon also took first place on Zapier's AutomationBench at 56.3%, ranked second on the Vals Index and second for steerability, and placed third on the Artificial Analysis Intelligence Index with a score of 61. Gemini 4 Neon takes second place on Agent Arena Index, trailing only Claude Fable 5.2 Photo: Tim Witzdam / Pexels Google DeepMind's newest model scored 1336 on Arena.ai's agent leaderboard while undercutting rivals on cost per task Google https://cryptobriefing.com/markets/alphabet/ DeepMind’s newest AI model, Gemini 4 Neon, has landed in second place overall on Arena.ai’s new Agent Arena Index. It posted a score of 1336, drawn from more than 4,200 real-world agent sessions. The only model ahead of it is Anthropic https://cryptobriefing.com/markets/anthropic/ ’s Claude Fable 5.2 Max , which scored 1463. What Gemini 4 Neon actually scored Google DeepMind launched Gemini 4 Neon on November 18, 2026. The company pitched it at software engineering, knowledge work, and science workflows: the kind of jobs where a model has to plan, act, check its work, and keep going. The Agent Arena Index measures how models perform in that setting. Neon’s 1336 came from over 4,200 sessions involving real-world agent work, not a single scripted test. The model also ranked second on the index for steerability. That refers to how well a model adjusts when a user changes direction partway through a task. Neon’s showing extends beyond Arena.ai. It took first place on Zapier’s AutomationBench with a score of 56.3%. AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. It ranked second on the Vals Index. On the Artificial Analysis Intelligence Index, it placed third with a score of 61. The cost argument Where Neon stands out most is the bill. On the Agent Arena Index, it posted a median cost of $2.82 per task. That makes it 53% cheaper than its nearest competitor, Claude Opus 5.5 Max . It is also 69% cheaper than the top-ranked model on the index. For developers paying by usage, Google lists Neon at $3 per million input tokens and $15 per million output tokens. A fast-moving Gemini 4 lineup Neon is the fourth release in the Gemini 4 series within seven weeks. Earlier entries in the family included Gemini 4 Carbon and Gemini 4 Hydrogen. On the same index, Carbon scored in the 1300s while Hydrogen sat in the low 1200s. Neon’s 1336 puts it ahead of those siblings. Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .