Google's Gemini 4 Neon arrives to challenge Claude and OpenAI in the AI mid-range

Source: Date:

Introduction: Google has officially entered the next phase of its AI model strategy with the launch of Gemini 4 Neon, a mid-tier model designed to strike a balance between raw capability and cost efficiency. Positioned squarely between the flagship Gemini 4 Carbon and the budget-friendly Gemini 4 Hydrogen, Neon is Google's latest attempt to capture developers, researchers, and enterprises who need frontier-level performance without paying premium prices. With competitive benchmark results that put it head-to-head against Anthropic's Claude Opus 5.5 and OpenAI's GPT-6.2 Sol, Neon could reshape how teams choose their everyday AI workhorse.

Google's announcement places significant emphasis on what Neon can do across software engineering, knowledge work, and scientific workflows. The company claims the model delivers frontier-level intelligence with fast end-to-end task completion, making it a versatile option for a broad range of professional applications.

When stacked against Claude Opus 5.5 and GPT-6.2 Sol, Neon holds its own in Google's published benchmarks. It comes out ahead on the Artificial Analysis Intelligence Index, scoring 61 compared to 59 and 57 for its rivals, and posts 70.9% on the Vals Index versus 70.1% and 66.1%. However, Anthropic's Opus 5.5 retains an edge on Terminal-Bench 4.1 and FrontierCode 2.0, suggesting the competition remains tight in specialized coding tasks. Google also highlights that Neon achieves its lowest hallucination rate to date in internal testing.

The model lineup is now clearly stratified. Gemini 4 Carbon targets the most demanding long-horizon projects and deep research initiatives, while Gemini 4 Hydrogen serves as the speed-focused, economical choice for everyday coding and agentic workflows. Neon occupies the middle ground, balancing depth and efficiency. Pricing reflects this hierarchy: Carbon costs $4 per million input tokens and $20 per million output tokens, Neon runs $3/$15, and Hydrogen is priced at $1.50/$7.50, though certain promotional rates may apply.

On the technical side, Neon features a 1 million-token context window and four configurable thinking levels, topping out at xhigh. Google states that at its default high thinking setting, Neon completes tasks at 32% lower cost than Carbon. The model can also function as an orchestrator, coordinating parallel Hydrogen subagents within Google Antigravity for complex multi-agent workflows.

Availability is broad from day one. Gemini 4 Neon is rolling out through the Gemini API in Google AI Studio, Gemini for business, and Google Antigravity for paid Google AI Pro and Ultra subscribers. Pro and Ultra members can also access it directly via the model selector in the Gemini app. Additionally, a specialized Neon for cyber variant will be distributed to vetted security defenders through Google's Fairwind Program in the coming days.

Source

Scroll to Top