All posts
2 min readby Romiel Inolino

Gemini 4 Neon costs $3 in and $15 out per million tokens and tops Zapier's AutomationBench at 56.3%, though it uses more output tokens per task

GeminiAI modelsLLM pricingautomationAI agents

Gemini 4 Neon is worth testing for automation workloads, but the per-task numbers matter more than the headline price.

What happened

Google released Gemini 4 Neon this week. It has a one million token context window and is priced at $3 per million input tokens, $0.15 cached, and $15 per million output tokens. It's available in the Gemini API through Google AI Studio and in Gemini for business.

Google says Neon scores 61 on the Artificial Analysis Intelligence Index at its xhigh level, third overall, and ranks first on Zapier's AutomationBench at 56.3% and on the Artificial Analysis Coding Agent Index. Google also claims its lowest fabrication rate yet in internal Antigravity testing and a 32% lower cost per task than Carbon at the default high level.

The Decoder adds useful detail. Neon's list price is 25% below Carbon's regular price but 50% above Carbon's promotional price, which runs through the end of November. Neon uses about 15% more output tokens per task than Carbon, so the savings come from the lower per-token price. It runs at roughly 110 tokens per second, close to double Carbon's 57.

My take

The AutomationBench result is the number I'd pay attention to, because it's Zapier's benchmark for business automation tasks, which is closer to my clients' work than math or coding scores. Still, 56.3% means the best model on that test fails a large share of tasks. Any automation built on it needs validation and a fallback path.

The token detail matters for budgeting. If Neon writes more output per task, a workflow that runs thousands of times a day won't see the full price cut. Run it on a sample of your real jobs and compare cost per completed task, not cost per token.

Teams currently on Carbon's promotional price should also check the math before December, since Neon may cost more for them until that promotion ends.

The speed gain is the clearest win. For agents that sit in front of a customer or a sales rep, halving response time matters more than a point on a benchmark.

More posts