Incident sheet · /enterprise-ai
- System
- Enterprise AI
- Filed
- Read
- 4 min · 752 words
- Severity
- Routine
Google Ships Gemini 4 Neon: $3 Per Million Input Tokens, 1M-Token Context
Google launched Gemini 4 Neon on November 18, 2026, priced at $3 per million input tokens with a 1M-token context window, positioned between Hydrogen and Carbon for coding and agent work.

Event log
- Google launched Gemini 4 Neon on November 18, 2026.
- API pricing is $3 per million input tokens, $0.15 cached, $15 output.
- Unified context limit is 1,048,576 tokens across input and output.
- Google claims 32% lower cost per task than Carbon at default High setting.
- Neon supports text, images, video, audio, and PDFs with four thinking levels.
Google launched Gemini 4 Neon on November 18, 2026, with API pricing set at $3 per million input tokens and $15 per million output tokens. The model targets demanding coding, analysis, science, and agent work that needs more reasoning than Google's speed-focused Hydrogen model.
Google positions Neon as a balance of reasoning depth, speed, and cost. It slots between Hydrogen, the fast everyday model, and Carbon, the model for complex long-horizon planning. Google says Neon is adapted from Gemini 4 Carbon.
What are the core specifications?
Google documents a unified context limit of 1,048,576 tokens shared between input and output. Tokens are pieces of text or other encoded content, so the exact word count varies. A large input leaves less room for the generated response.
Neon accepts five input types: text, images, video, audio, and PDFs. Google lists multi-file software engineering, analytical reasoning, STEM work, enterprise synthesis across documents, and long-running tool-heavy agent workflows as intended use cases. In agent workflows, a system plans several steps, calls external tools, and reviews results rather than producing one immediate answer.
The model offers four thinking settings: low, medium, high, and xhigh. High is the default. Standard synchronous responses are capped at 65,536 output tokens. Google also documents long decode up to 1,048,576 tokens, subject to the unified context limit.
What does the cost claim mean?
Google Antigravity reports that Neon's cost per task is 32% lower than Carbon's at the default High thinking setting. That figure describes Google's comparison under its stated conditions only. Actual results vary with task length, tool calls, retries, and thinking settings.
On the paid Gemini API tier, cached input is priced at $0.15 per million tokens. Google describes that as a 95% discount versus the standard input rate. These are usage charges, not subscription prices; API bills depend on tokens processed.
Where is Neon available?
Google lists five access routes:
- Developers can use Neon through the paid Gemini API tier, including Google AI Studio.
- Google AI Pro and Ultra subscribers can access Neon in AI Studio and the Gemini app, subject to rollout.
- Gemini for business is another listed route.
- In Google Antigravity, Neon is available to paid Pro and Ultra subscribers.
- Developers can also try Neon through Android Studio.
Availability varies by country, product, and plan. A subscription does not guarantee immediate access in every listed app or account.
How should enterprises evaluate it?
Google describes Neon as strong in coding and STEM, but benchmark claims require context. A score means little without the benchmark name, test date, model settings, and whether tools or multiple attempts were allowed. Google's results are not an independent head-to-head test, and scores from different benchmarks measure different tasks, so they should not be treated as a single ranking.
Google Antigravity's 32% cost-per-task figure does not promise lower costs for every prompt or pricing setup, and a model that uses fewer tokens may still take longer or require more corrections, which is why running a small batch of representative tasks before switching matters.
Practical testing guidance follows the same logic. Run identical prompts through Neon and the current model with comparable settings and tools. Check outputs against trusted sources. Note errors, omissions, and corrections. Compare response time and token use as well as answer quality, because a longer response or extra tool calls can change total cost.
Who is the model for?
Neon fits tasks that need more than a quick answer, especially long inputs, mixed media, or multiple tool calls. Developers could test it on large-codebase reviews tracing how a change affects several files. Researchers could submit long reports for comparison or theme identification. Teams could combine a written brief with a chart or a recording.
Hydrogen remains the better fit for short routine requests. Carbon suits work needing deeper planning. Neon is a reasonable starting point for coding, analysis, or STEM tasks that sit between those needs.
Two caveats apply. First, Google's model positioning is product guidance, not proof of superiority on every task. Second, the Gemini API data logging policy and organizational rules should be reviewed before testing. Google advises against submitting private, customer, employee, or otherwise sensitive information unless workplace policies and service terms allow it.
Neon is one model within the Gemini 4 family, not a separate consumer app. Google also offers a separate family called Gemma 4, which is not another name for Neon.
via chiangraitimes.com (Source)
More from Rebecca Stone
Show full bio
Senior reporter covering money, funding and market shifts around enterprise information technology and the leaders who run at CIO Monitor.
208 articles
The newsroom digest
One email a day with the stories that matter. No noise.
Cross-references
- Google rolls out Gemini 4 Neon, citing 32% lower task cost
- Google Ships Gemini 4 Hydrogen to All Search Users Globally
- Google ships Gemini 4 Argon; intro pricing doubles after November 30
- Google's Gemini Decisions API Scores 1,000 Questions Per Call
- Google to lock Gemini Pro behind $19.99 AI Pro plan from October 9