Incident sheet · /enterprise-ai
- System
- Enterprise AI
- Filed
- Read
- 3 min · 591 words
- Severity
- Routine
Google rolls out Gemini 4 Neon, citing 32% lower task cost
Google’s Gemini 4 Neon combines a $3-per-million input price with a one-million-token context window, while researchers say production code still needs human oversight.

Event log
- API pricing is $3 per million input tokens, $15 per million output tokens and $0.15 per million cached-input tokens.
- Neon offers four reasoning levels and a context window of up to one million tokens.
- Neon scored 58.6 on the Artificial Analysis Coding Agent Index and 65.2% on Terminal-Bench 4.1.
- Google reports 32% lower cost per task than Gemini 4 Carbon at the default high reasoning setting.
- Google reports roughly 50% to 80% savings against leading models on the Vals Index and Agent Arena benchmarks.
By Olivia Hart
Google says Gemini 4 Neon cost 32% less per task than Gemini 4 Carbon in Artificial Analysis Intelligence Index evaluations, with API pricing set at $3 per million input tokens and $15 per million output tokens.
For enterprise developers, that puts a lower-cost option between two existing Google models and supports Google’s push to expand AI coding agents among enterprise teams. Neon handles complex software engineering work across large code repositories, including multi-file debugging, refactoring and targeted changes to existing code.
Inside Google’s Antigravity development environment, Neon can also orchestrate multiple Gemini 4 Hydrogen subagents in parallel. The model supports a context window of up to one million tokens.
What does this change for enterprise development teams?
Google positions Neon between Gemini 4 Carbon, designed for demanding projects, and Gemini 4 Hydrogen, which prioritizes faster execution. Developers can select from four reasoning levels to adjust computational effort to each workload.
The production constraint starts with visibility. Anand Joshi, managing director of market intelligence firm JP Data, said: “The challenge with AI-generated code is that it could be a black box to the developer.”
Without a reliable way to understand generated code, developers may struggle to identify problems when applications fail in production. Joshi said that uncertainty could keep enterprises from moving beyond prototyping.
Google reported these launch metrics:
- Neon ranked first on the Artificial Analysis Coding Agent Index, with a score of 58.6.
- It achieved 65.2% on Terminal-Bench 4.1, a benchmark for AI agents performing tasks in terminal environments.
- At the default high reasoning setting, it delivered 32% lower cost per task than Carbon in Artificial Analysis Intelligence Index evaluations.
- On the Vals Index and Agent Arena benchmarks, Google reported savings of roughly 50% to 80% against leading competing models.
Cached Neon inputs cost $0.15 per million tokens. Google is making the model available through the Gemini API in Google AI Studio and through Gemini for Business, with access also expanding to the Antigravity development platform.
The immediate enterprise effect is a lower-cost middle tier for coding-agent work. Production software often needs long context, which can push token consumption beyond initial estimates, Joshi said. Smaller, more efficient models could draw buyers, provided they deliver code of comparable quality.
Prabhu Ram, vice president for industry research at CyberMedia Research, expects a tiered approach rather than wholesale abandonment of premium models. Enterprises should choose each system according to the requirements and economics of the development task.
Ram also urged buyers to measure the total cost of successfully completing a task, not benchmark scores alone. Code quality and compatibility with existing development environments should shape purchasing decisions.
Autonomous agents may not yet be ready to manage production software independently because reliability and the complexity of enterprise codebases remain significant obstacles. Ram said: “The near-term opportunity lies in supervised autonomy, with AI agents handling increasingly complex tasks while developers retain oversight and accountability.”
Pareekh Jain, CEO of Pareekh Consulting, warned that coordinating multiple coding agents can introduce conflicting changes and errors. Long-running development tasks also create reliability problems, particularly for business-critical software.
Enterprises should test generated applications against normal and edge-case scenarios, including end-to-end and regression testing. Joshi said functional testing will become a key requirement alongside security and cost. Jain also recommends measuring developer productivity gains and the human intervention needed to complete tasks reliably.
Neon lowers the price of AI-assisted development, but enterprises still have to own the code it writes.
via infoworld.com/feed/ (Source)
More from Olivia Hart
Show full bio
News editor covering money, funding and market shifts around enterprise information technology and the leaders who run at CIO Monitor.
207 articles
The newsroom digest
One email a day with the stories that matter. No noise.
Cross-references
- Google Ships Gemini 4 Neon: $3 Per Million Input Tokens, 1M-Token Context
- Google Ships Gemini 4 Hydrogen to All Search Users Globally
- Google ships Gemini 4 Argon; intro pricing doubles after November 30
- Google's Antigravity Orchestrator Adds Gemini 4 Hydrogen
- Google to lock Gemini Pro behind $19.99 AI Pro plan from October 9