On November 18 Google launched Gemini 4 Neon (gemini-4-neon), positioned between Gemini 4 Hydrogen and Carbon and promising frontier-level intelligence with faster end-to-end task completion. Google says it scores 61 on the Artificial Analysis Intelligence Index at xhigh thinking (#3 overall), 70.9% on the Vals Index (#2), 58.6 on the Artificial Analysis Coding Agent Index (#1) and 65.2% on Terminal-Bench 4.1. It costs $3 per million input tokens and $15 per million output tokens ($0.15 for cached input) with a 1M-token context. It is rolling out in the Gemini API and AI Studio, Gemini for business, Google Antigravity for paid AI Pro and Ultra subscribers and the Gemini app for Pro and Ultra, and Vertex AI pricing is live.
Key Takeaways
- ✓Positioning: between Hydrogen (the everyday coding default) and Carbon (the hardest architecture, debugging and security problems), aimed at high-frequency complex engineering, knowledge work and science
- ✓Benchmarks (Google): 61 on the AA Intelligence Index at xhigh (#3 overall), 70.9% on the Vals Index (#2), 58.6 on the AA Coding Agent Index (#1), 65.2% on Terminal-Bench 4.1, 65.4% on CursorBench 4.1 and 56.3% on AutomationBench
- ✓Pricing: $3 per million input tokens and $15 per million output tokens including thinking; cached input is 95% cheaper at $0.15; Google says cost per task at high is 32% lower than Carbon
- ✓Specs: a unified 1,048,576-token context, a 65,536-token default synchronous output cap and long decode up to 1M with background execution; low, medium, high and xhigh thinking, default high
- ✓Availability: the Gemini API and AI Studio, Gemini for business, Vertex AI, Google Antigravity for paid AI Pro and Ultra subscribers, and the Gemini app for Pro and Ultra subscribers
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background
The Gemini 4 lineup had two clear ends: Hydrogen, cheap and fast and the everyday coding default, and Carbon, the most capable but slower and pricier. On November 18 Google launched Gemini 4 Neon to fill the middle tier. The official Google DeepMind X account says it sits between Carbon and Hydrogen.
How it works
Google says Neon targets faster end-to-end task completion: more surgical edits in repository-level debugging and multi-file refactors, fewer redundant file reads in long sessions, and its lowest fabrication rate yet in internal Antigravity testing. It has a unified 1M-token context, a 65,536-token synchronous output cap and long decode up to 1M with background execution, with four thinking levels defaulting to high.
Numbers that matter
Per Google's blog and model card: 61 on the AA Intelligence Index at xhigh (#3 overall) versus 59 for Claude Opus 5.5 and 57 for GPT-6.2 Sol; 70.9% on the Vals Index (#2); 58.6 on the AA Coding Agent Index (#1); and 65.2% on Terminal-Bench 4.1, just below Opus 5.5 at 65.8%. Pricing is $3 per million input tokens and $15 per million output tokens.
What it means for developers
On the paid tier, set the model ID to gemini-4-neon; there is no free tier. Explicit caching is unsupported and temperature, top_p and top_k are ignored. AI Pro and Ultra subscribers can also use it in Antigravity and the Gemini app.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.