Skip to content
Theme
OnDevice Tools

AIChatbots & Assistants

Google launches Gemini 4 Neon mid-tier AI model

Google says Gemini 4 Neon balances capability and cost, with a one million-token context window and lower per-task pricing than Gemini 4 Carbon.

Illustration of Google Gemini 4 Neon AI model with benchmark charts and token pricing

Key takeaways

  • Google launched Gemini 4 Neon as a mid-tier model in the Gemini 4 family.
  • Google says Neon scores 61 on the Artificial Analysis Intelligence Index and 70.9% on the Vals Index.
  • Neon costs $3 for input and $15 for output per million tokens, below Gemini 4 Carbon.
  • The model has a one million-token context window, four thinking levels and can coordinate Hydrogen subagents.
In this article
  1. What Google announced
  2. Benchmark results and pricing
  3. Where Neon fits in the Gemini 4 lineup
  4. What it means for developers

What Google announced

According to GSMArena, Google has launched Gemini 4 Neon. It is the newest addition to the Gemini 4 family of artificial intelligence models.

Google describes Neon as a mid-tier model. It says the model aims to combine strong reasoning with fast completion of multi-step tasks.

The company says Neon is suited to coding, professional work and science projects. In plain terms, that covers jobs such as writing software, handling research and supporting technical analysis.

The model is not presented as a replacement for Google's largest or smallest models. Instead, it is pitched as a practical option for many tasks. The announcement is based on Google's own materials, and the benchmark results have not been independently confirmed in the source report.

Benchmark results and pricing

Google's comparison places Neon against Claude Opus 5.5 and GPT-6.2 Sol. On the Artificial Analysis Intelligence Index, Neon scores 61. The rivals score 59 and 57. On the Vals Index, Neon scores 70.9%, compared with 70.1% and 66.1%.

The scores suggest Neon is near the top of the tested group. However, the ranking depends on the task. The results are not a clean sweep.

Google's own chart shows Claude Opus 5.5 ahead on Terminal-Bench 4.1 and FrontierCode 2.0. The report does not give detailed methodology for these tests. This means the overall ranking depends on which benchmark is used.

Google also says Neon produced fewer false statements in internal testing than previous models. The report does not include the full test data. This claim should be treated as a vendor statement rather than an independent result.

Pricing is $3 per million input tokens and $15 per million output tokens. Tokens are the small units of text that AI models read and generate.

That places Neon below Gemini 4 Carbon, which costs $4 and $20. It also places Neon above Gemini 4 Hydrogen, which costs $1.50 and $7.50. Google says some promotions may apply to some models. The price gap may matter for high-volume users.

Where Neon fits in the Gemini 4 lineup

Illustration: Gemini logo, AI, environment, Google logo and the labels Gemini 4, Gemini 4 Neon, Google

Neon is positioned between Gemini 4 Carbon and Gemini 4 Hydrogen. Carbon is intended for the hardest long-horizon projects, meaning complex work that unfolds over many steps. It is also aimed at deep research. Hydrogen is the fastest and cheapest option for everyday coding and agentic workflows, where AI systems carry out multi-step tasks.

Neon is pitched as a balance between capability and cost. It can handle a context window of one million tokens. That means it can process a very large amount of text in a single request. It also offers four thinking levels, with xhigh at the top.

At its default high setting, Google says Neon is 32% cheaper per task than Carbon. Neon can also coordinate parallel Hydrogen subagents inside Google Antigravity. This allows several smaller agents to work at the same time.

These features support work that needs long inputs or careful reasoning. The orchestrator role also points to multi-agent workflows. In such setups, one model directs other models to complete parts of a job.

What it means for developers

For developers and businesses, Neon adds a middle choice in Google's AI product range. It may suit applications that need more capability than Hydrogen but lower cost than Carbon. Possible uses include coding assistants, research tools and internal knowledge systems.

The key limitation is that the evidence comes from Google. The benchmark comparison names rival products. However, the source report does not provide independent testing. Organisations should review official documentation and run their own evaluations before deployment.

The launch also highlights a wider trend. AI vendors are splitting model families by price, speed and complexity. Neon's role is to cover the middle ground. Carbon targets the most demanding projects, while Hydrogen handles routine, high-volume work.

The model's success will depend on real-world performance. Google's benchmarks provide one signal, but they are not the whole picture.

Sources

  1. Google launches Gemini 4 Neon, its new mid-tier AI modelGSMArena

Tags

Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.