Google Introduces Gemini 4 Neon for Coding and Enterprise Workflows

Google has introduced Gemini 4 Neon, a new model with a million-token context window for programming, analytical tasks and enterprise workflows. Deployment is beginning in the API, business products and paid AI plans.

On November 18, Google announced Gemini 4 Neon, a new model in the Gemini 4 family designed for complex programming, analytical tasks and enterprise workflows. The company places it between the faster Gemini 4 Hydrogen and the more powerful Gemini 4 Carbon. The model is beginning to roll out through the Gemini API in Google AI Studio, Gemini for business, and to Google AI Pro and Ultra subscribers in the Gemini app and Google Antigravity.

The new model enters the segment of powerful language models designed not only to generate text, but also to work with code and multi-step workflows. Alongside the announcement, Google published a model card with its own evaluation results, a description of intended use and known limitations.

Gemini 4 Neon Offers a Million-Token Context

One of the model’s stated parameters is a context window with a capacity of 1 million tokens. This scale makes it possible to process a large amount of input material within a single request, such as extensive source code or documentation.

Google set the price of the paid API at $3 per million input tokens and $15 per million output tokens. These are the manufacturer’s listed prices and should be distinguished from the total costs of real-world deployment, which may vary depending on the type of task and the amount of output generated.

Google reports the results of its own benchmarks and evaluations, but complete independent verification of all stated comparisons is not yet available. Claims about quality, the cost of a specific task or the model’s position relative to competing models therefore remain preliminary until confirmed by external tests and practical comparisons.

Availability Is Rolling Out Gradually

According to Google, the model is available through the Gemini API in Google AI Studio. The rollout also covers Gemini for business and the paid Google AI Pro and Ultra plans in the Gemini app and Google Antigravity. However, the announcement does not specify the extent to which the model will be accessible during the rollout in individual countries, for different accounts and plans.

For developers and companies, the key factors are the stated focus on coding and agentic workflows, the large context window and the published API pricing. Google is thereby expanding its commercial offering of models designed for technical and enterprise tasks.

Cybersecurity Version Limited to Trusted Defenders

Google also said that a version of Gemini 4 Neon designed for cyber defense will be available in the coming days only to trusted defenders in the Fairwind program. The published information does not provide details about the program’s conditions or the scope of security testing for this version.

The announcement also does not confirm any specific exploitation of a vulnerability. The cyber variant is presented as a tool for defensive work, not as a response to an identified incident.

Model Card Warns of Limitations

In the model card, Google also lists limitations, including the possibility of hallucinations—cases in which the model produces an incorrect or unsupported response. Outputs therefore need to be verified when working with generated code, analyses or enterprise decisions.

Gemini 4 Neon’s knowledge cutoff is March 2026. The model may therefore not know information and events after that date unless they are provided directly in the prompt or through connected tools.

Further developments will be shown by independent benchmarks and practical comparisons with Claude Opus 5.5, GPT-6.2 Sol and Gemini 4 Carbon. It will also be important to see how quickly and to what extent the new model appears in the API, the Gemini app and business products, as well as any updates to Google’s model card and evaluation methodology.

Sources

  • Google Blog – Confirms the announcement, the model’s purpose, pricing, context window and stated availability channels.
  • Google DeepMind — Gemini 4 Neon Model Card – Confirms evaluation results, known limitations, the knowledge cutoff and intended use, including defensive cyber work.
  • 9to5Google – Independently corroborates the model’s launch and rollout for paid Google AI plans.

Verified and updated: November 19, 2026 06:24

Sharing