
Google released Gemini 4 Neon on 18 November 2026, a mid-tier model in its Gemini 4 family that it prices at $3 per million input tokens and $15 per million output tokens. Google’s model card reports results on science and math benchmarks, and the model is available through the Gemini API, Google AI Studio, Google Antigravity and GitHub Copilot, which matters to labs that pay for model access by the token.
What happened
According to 9to5Google, which reported the launch on 18 November, Google is rolling out Gemini 4 Neon after releasing Carbon and Hydrogen the previous week. Google describes Neon as offering “frontier-level reasoning for complex, high-frequency workflows faster and more cost efficiently.” Neon replaces Argon in the lineup, and in the Gemini app it is available to AI Pro and AI Ultra subscribers.
The Gemini 4 Neon model card, published on 18 November, says the model is adapted from Gemini 4 Carbon and is built for complex coding, analytical reasoning and production-grade enterprise workflows. It accepts text, images, audio and video, with a combined input-output context window of up to one million tokens, and it produces up to 64,000 tokens by default in a synchronous response or up to one million through background long decode.
Key details
The card lists the distribution channels as the Gemini app, the Gemini API, Google AI Studio, Gemini for business and Google Antigravity. It adds that access varies by channel and subscription tier. 9to5Google reports the model keeps four thinking levels (Low, Medium, High and Extra High) and the same one million-token context window as Hydrogen.
| Item | Detail | Source |
|---|---|---|
| Neon price | $3 per million input tokens, $15 per million output tokens; twice the price of Hydrogen | 9to5Google |
| Argon and Carbon price | $2 and $10 until 30 November, then $4 and $20, which makes Neon 25% cheaper after that date | 9to5Google |
| Knowledge cutoff | March 2026, with some domains less current | Model card |
| Context window | Up to one million tokens, input and output combined | Model card |
| Copilot availability | Copilot Pro+, Max, Business and Enterprise, rolling out gradually | GitHub changelog |
Google’s model card compares Neon with Claude Opus 5.5 and GPT-6.2 Sol. On the science, math and reasoning rows, the card reports the following results as of November 2026. These are Google’s own published figures, and the card states that its safety evaluation results are automated evaluations and not human evaluation or red teaming.
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| RiemannBench (science and math) | 84.0% | 80.0% | 80.0% |
| Terminal-Bench-Science 0.1 | 70.5% | 63.3% | 68.0% |
| HLE-Diamond (reasoning) | 71.3% | 67.9% | 64.0% |
| AA Index v5.0.1 (capability index) | 61 | 59 | 57 |
Background and context
The launch follows the Gemini 4 releases earlier in the month that casrai covered, including Gemini 4 Carbon and Gemini 4 Hydrogen. 9to5Google places Neon between Hydrogen, which it describes as a Flash replacement for everyday help, and Carbon, which Google positions for the most complex work. It also reports that Gemini 3.1 Pro is going away from the Gemini app.
Google’s Antigravity team said on 18 November that it is rolling out Neon inside Google Antigravity, and that Orchestrator now supports pairing Neon as the planner with Hydrogen as the executor in one session. The team reports that at its default High thinking level Neon shows a 32% lower cost per task than Carbon in its testing, and describes Neon’s lowest fabrication rate yet in internal Antigravity testing.
Reaction
GitHub added the model to Copilot the same day. In its changelog entry, GitHub says that, according to Google, Neon delivers quality close to Carbon at a lower token cost per task, starts responding faster and makes fewer unrequested changes on coding tasks. GitHub also relays Google’s statement that the model is less likely to make unsupported claims about code, tool or test results. GitHub notes the model is billed at provider list pricing under usage-based billing.
The model card is itself cautious on limits. It says Gemini 4 Neon may exhibit general limitations such as hallucinations, that Google is continuing to improve jailbreak resistance, and that the full Safety and Security Report will be available in the coming days.
What it means for researchers, universities and research offices
For groups that run analysis pipelines or coding agents against a paid API, the published prices make Neon the middle option: $3 and $15 per million tokens against $4 and $20 for Carbon once the promotional pricing ends on 30 November. Research offices that budget grant-funded compute by the token should compare cost per task rather than list price alone, since Google’s Antigravity team reports a lower cost per task than Carbon at the default thinking level.
The science and math scores come from Google’s card, so labs relying on them for method choices should treat them as vendor-reported until independent evaluators publish results. Casrai’s guide to choosing and governing large language models for research sets out a framework for that comparison, and the entry on hallucination explains the failure the card lists among Neon’s known limitations. The knowledge cutoff of March 2026 matters for literature-review tasks.
For IT and research-computing administrators, GitHub’s changelog sets out a concrete step. Copilot Enterprise and Copilot Business administrators manage access through the model policy in Copilot settings, and under default model enablement new models are switched on automatically unless an administrator has turned off the global default or explicitly disables the model. Institutions with data-handling rules for unpublished results should check that setting. The card also says that access through the Gemini API is subject to the Gemini API Additional Terms of Service and access through Gemini for business to the Google Cloud Platform Terms, which research-data offices should read before sending protected data. Background on frontier models is in casrai’s guide to what counts as a frontier AI model.
What happens next
The model card says the full Safety and Security Report for Gemini 4 Neon will be published in the coming days, and the card’s safety section points to the Gemini 4 Carbon report for evaluation approach and safety policies. The Argon and Carbon promotional pricing ends on 30 November, according to 9to5Google. GitHub says the Copilot rollout is gradual, and the Gemini app picker now shows Neon under the deeper analysis option.
Sources
- Google DeepMind: Gemini 4 Neon model card, published 18 November 2026, checked 2026-11-19
- 9to5Google: Gemini 4 Neon launches: Carbon quality at Flash speeds, for AI Pro, 18 November 2026, checked 2026-11-19
- GitHub Changelog: Gemini 4 Neon in GitHub Copilot, 18 November 2026, checked 2026-11-19
- Google Antigravity Blog: Gemini 4 Neon in Google Antigravity, 18 November 2026, checked 2026-11-19







