What Just Changed
Gemini 4 Neon is now live for AI Pro and AI Ultra subscribers in the Gemini app. It replaces Argon in the model lineup and slots neatly between Hydrogen (everyday help) and Carbon (the heavy lifter). The updated model picker now looks like this:
- Auto — Adapts to your needs
- Everyday help — Gemini 4 Hydrogen
- Deeper analysis — Gemini 4 Neon
- Code & complex work — Gemini 4 Carbon
3.1 Pro is also officially gone. And 3.5 Flash-Lite has been quietly renamed “Auto” for paid users — a small but telling signal about where Google sees the baseline.
What Neon Actually Does
The pitch is Carbon-level quality at something closer to Flash speeds. That’s not just marketing framing — it’s a real positioning choice that shapes what Neon is built for.
It carries the same 1 million-token context window as Hydrogen, with four thinking levels (Low, Medium, High, Extra High). That’s a lot of room to work with for document-heavy tasks.
Where it’s designed to shine:
- Reading and summarizing dense uploads — PDFs, spreadsheets, photos
- Research tasks that need current web sources baked in
- Multi-step agentic tasks that would burn through usage limits on Carbon
- Pulling context from Gmail and Docs for knowledge work
It’s not trying to replace Carbon for your hardest long-horizon projects. It’s trying to handle the 80% of complex work that doesn’t need Carbon’s full weight.
The Benchmark Picture
Benchmark numbers should always be read with some skepticism, but Neon’s results are worth noting across a few specific categories.
On the Artificial Analysis Intelligence Index, Neon scores 61 at its Extra High thinking level — ranking third overall, behind only larger, slower models. It reaches first place on the Legal Research Bench, Tax Agent Bench, and the Artificial Analysis Coding Agent Index. It also debuts second globally on Agent Arena’s new index across 4,200+ real-world agent sessions.
For automation-heavy use cases, it takes first on Zapier’s Automation Bench (56.3%) and Andon Labs’ Vending-Bench 2 — a long-horizon simulation of running a business end-to-end. That last one is a useful proxy for anyone building or evaluating agentic workflows.
It scores 73.5% on Chartography (reading dense professional charts) and 75.8% on Agents’ Last Exam v1.1 — roughly on par with leading models.
Pricing: The Part That Actually Matters
Neon is priced at $3 per million input tokens and $15 per million output tokens — twice the cost of Hydrogen, but 25% cheaper than Carbon once promotional pricing ends.
For context, Argon and Carbon are currently available at $2/1M input and $10/1M output through November 30. After that, they move to $4/1M input and $20/1M output — which is where Neon’s pricing advantage becomes concrete.
If you’re running production workflows at scale, Neon’s positioning as the efficient mid-tier starts to make real financial sense, especially for tasks that don’t require Carbon’s ceiling.
Who Should Pay Attention
If you’re an AI Pro or Ultra subscriber doing any of the following, Neon is worth switching to immediately:
- Analyzing documents, contracts, or financial statements
- Running research briefs with live web context
- Building or testing multi-step agents
- Coding tasks that don’t need Carbon’s full reasoning depth
If you’re evaluating models for enterprise workflows, Neon’s benchmark wins in finance, accounting, spreadsheet modeling, and constraint-following suggest it’s not just a consumer upgrade — it’s a serious contender for production use.
The Useful Takeaway
Neon isn’t trying to be the smartest model in the room. It’s trying to be the most useful one for the work most people actually do — complex enough to handle real tasks, efficient enough not to burn your quota on every query.
For AI Pro subscribers, the practical move is simple: swap Argon for Neon, run your usual workflows, and see if the quality gap with Carbon is worth the cost difference for your specific use case. Odds are, for most tasks, it won’t be.
Comments (0) No comments yet
Want to join this discussion? Login or Register.
No comments yet. Be the first to share your thoughts!