Gemini 4 Neon is live in India — and it’s already replacing Gemini 4 Argon in the Gemini app. Google confirmed the launch on November 18, 2026, positioning Neon as the new default for complex reasoning tasks that don’t need Carbon’s full heft. For Indian users on Google AI Pro, it’s accessible today under “Deeper analysis”, no extra fee beyond the Rs 250/month subscription.
Google AI Ultra users get it too. Both plans include unlimited access to Neon across text, audio, image, and video inputs. Developers in India can tap it now via Google AI Studio, Android Studio, Gemini for business, and Google Antigravity, with a cybersecurity-specific version slated for Fairwind Program members soon.
There is no separate Indian pricing for API access: the global rate applies, $3 per million input tokens, 15 cents per cached input, and $15 per million output tokens. That puts Neon squarely between OpenAI’s GPT-6.2 Sol ($2 input) and Anthropic’s Claude Opus 5.5 ($4 input). But there’s a catch this month: Gemini 4 Carbon remains discounted at $2 input and $10 output until November 30.
From December 1, Carbon jumps to $4 and $20, making Neon the more cost-efficient choice for speed-sensitive workloads. Google says Neon runs roughly three times faster than Carbon while cutting inference costs by about 60 per cent. Independent testing backs the speed claim: Artificial Analysis gave Neon a 61 on its Intelligence Index at Extra High thinking level, just behind GPT-6.2 Astra (62) and Claude Fable 5.2 (65), and ahead of Carbon’s 60.
In domain-specific benchmarks, Neon scored 70.5% on Terminal-Bench Science, 84% on RiemannBench, and 71.3% on HLE-Diamond. It topped both Legal Research Bench and Tax Agent Bench, hit 73.5% on Chartography, and earned $16,843 in Andon Labs’ Vending-Bench 2 simulation. On safety, it posted zero per cent success rate on Gray Swan’s prompt injection test, 90.5% on SEC-bench Pro for bug spotting, and 99.2% honesty in Google’s internal evaluations.
Neon supports four adjustable thinking levels, Low, Medium, High, and Extra High, letting developers trade off latency for depth. It does not replace Hydrogen (the lightweight daily-task model launched November 12) or Carbon (the flagship for massive-scale projects). Instead, it fills the gap: high-end reasoning for coding, research, and data analysis without waiting or overpaying.
Gemini 3.1 Pro is officially retired. The rollout is global and immediate, no regional delays. Indian developers and Pro subscribers don’t need to wait.
ADFiled by The AI Desk
Models, assistants and the companies and chips behind them, reported from what was released and what was claimed, with the difference kept clear.
More from this desk →
Be the first to comment
Join the argument. No password, just your email or a passkey.