Back to News
ProductAI Understanding briefing

Google launches Gemini 4 Neon, a mid‑tier model for faster coding with a million‑token context

Google unveiled Gemini 4 Neon, a mid‑tier AI model that promises three‑times faster coding tasks than its flagship Carbon while costing about 60% of the price, and adds a one‑million‑token context window.

5 min readRead the linked source
Source-provided image accompanying Google launches Gemini 4 Neon, a mid‑tier model for faster coding with a million‑token context
Source referenceSource recorded
Publisher
m.dailyhunt.in
Source type
Linked source — primary-source status has not been established.
ContextUnderstand this in 60 seconds

Key terms

Token
A chunk of text processed by language models, such as a word piece or symbol.
API (Application Programming Interface)
A structured way for one software system to send requests to and receive responses from another system.
Context Window
The maximum amount of input tokens a language model can process at once.

What happened

Google introduced Gemini 4 Neon, its fourth Gemini 4 variant in seven weeks, positioning it between the deep‑thinking Carbon and the speed‑focused Hydrogen models. Neon supports four reasoning settings (low, medium, high, extra‑high) and a one‑million‑ , enabling long‑running, tool‑intensive coding jobs and multimodal analysis of spreadsheets, PDFs, charts, audio and video. Google says Neon completes tasks roughly three times faster than Carbon at about 60% of Carbon’s cost, with a reported 32% lower cost per task at the default high setting. The model ranks third on Artificial Analysis’s Intelligence Index (score 61) and leads the Coding Agent Index (58.6). API pricing is $3 per million input tokens, $15 per million output tokens, with cached input at $0.15. Neon is being rolled out via AI Studio, Gemini for Business, and the Gemini app for Pro and Ultra subscribers, with a “cyber” version for trusted defenders through the Fairwind Program.

Google announced Gemini 4 Neon as the latest addition to its Gemini 4 family, following the October 1 launch of Argon and the earlier Gemini 3.1 Pro. Neon is described as a mid‑tier model that balances the depth of the flagship Carbon model with the speed of the Hydrogen model.

The model provides four configurable reasoning settings—low, medium, high, and extra‑high—allowing users to trade off usage against performance. It also introduces a one‑million‑token , enabling it to process much larger inputs than previous Gemini models.

Google claims Neon completes tasks about three times faster than Carbon while costing roughly 60% of Carbon’s price. At the default high setting, Google reports a 32% lower cost per task, though an independent analysis by Artificial Analysis estimates average savings of about 17%, indicating variability across workloads.

Neon scores 61 on Artificial Analysis’s Intelligence Index (third place) and leads the Coding Agent Index with a score of 58.6. It also ranks second on the Vals Index. Google says the model can handle long‑running, tool‑intensive coding jobs, coordinate parallel Hydrogen sub‑agents in its Antigravity framework, and analyze multimodal data such as spreadsheets, PDFs, charts, audio, and video.

API pricing is set at $3 per million input tokens and $15 per million output tokens, with cached input tokens priced at $0.15. These rates are higher than Carbon’s promotional pricing, so cost savings depend on Neon’s reduced consumption. The model is being rolled out through AI Studio, Gemini for Business, and the Gemini app for Pro and Ultra subscribers, with a specialized “cyber” version for trusted defenders via the Fairwind Program.

Source details: m.dailyhunt.in

Why it matters

Gemini 4 Neon represents a notable shift toward more cost‑effective, high‑throughput AI for software development and enterprise knowledge work. By offering a million‑ , the model can handle substantially larger codebases and multimodal data sets in a single prompt, reducing the need for chunking and manual orchestration. The claimed three‑fold speed improvement and lower per‑task cost could make AI‑assisted coding more accessible to mid‑size firms that previously found flagship models prohibitively expensive. Additionally, the tiered reasoning settings give developers granular control over token consumption, potentially translating into predictable budgeting for AI‑driven workflows. However, the pricing is higher per token than the promotional rates for Carbon, meaning actual savings depend on Neon’s ability to use fewer tokens—a factor that varies by workload. The model’s inclusion in Google’s Antigravity sub‑agent framework and its availability to Pro/Ultra subscribers also signals a tiered access strategy that may limit broader adoption until the model proves its efficiency gains in real‑world deployments.

The million‑ expands the practical limits of AI‑assisted coding, allowing developers to feed entire codebases or large documentation sets into a single prompt, which can streamline debugging, refactoring, and code generation workflows.

Speed and cost improvements are critical for enterprises that need to run large numbers of coding or data‑analysis tasks. If Neon delivers on its three‑fold speed claim, it could reduce compute time and associated cloud costs, making AI‑driven development more economically viable for a broader range of companies.

Tiered reasoning settings give organizations finer control over budgets, enabling predictable cost management—a key concern as AI usage scales across teams.

The model’s multimodal capabilities open possibilities for integrated analysis of business documents, spreadsheets, and media, potentially automating complex reporting and insight‑generation tasks that currently require manual effort.

However, the higher per‑ pricing relative to Carbon’s promotional rates means that actual savings are contingent on Neon’s efficiency. Without independent benchmarks, it remains uncertain how much token reduction users will achieve in typical workloads.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Thinking Budget (Test-Time Tokens):1,024 tokens
Complex Accuracy79%Math & Code Logic
Latency3.2sTime to first full output
Inference Cost$0.0092Per query estimated
Reasoning StyleStep VerificationInternal chain depth
Active Thinking Trace:
1Deconstruct user problem into formal constraints
2Propose candidate hypotheses & step-by-step calculation
3Self-correction: Backtrack and refute subtle edge cases
4Exhaustive consistency check & final output synthesis
Core takeaway: Test-time compute fundamentally changes AI economics. Instead of only scaling during pre-training, giving reasoning models more tokens at inference time allows them to systematically solve PhD-level STEM problems.
Interactive Concept Check+10 Points
AI Models Explained Quiz

Which component of an AI application is the machine-learning model itself?

What to watch next

Watch for independent benchmarks that compare Neon’s speed and cost claims against Carbon and other mid‑tier models, especially on large‑scale code generation and multimodal tasks. Monitor adoption rates among enterprise developers and any pricing adjustments Google may make if ‑usage savings do not materialize as advertised. Follow the rollout of the Fairwind “cyber” version to see how Google extends Neon’s capabilities to security‑focused use cases, and watch for any updates to the model’s safeguards or honesty test scores as it scales to broader user bases.

Independent performance and cost benchmarks comparing Neon to Carbon, Hydrogen, and competing models from other providers.

Adoption metrics among enterprise developers, especially in software engineering and data‑analytics teams, to gauge real‑world uptake.

Potential pricing adjustments or promotional offers from Google if Neon’s ‑efficiency does not meet expectations.

Updates to the Fairwind “cyber” version, including any new security‑focused features or expanded access beyond trusted defenders.

Future releases or enhancements to Neon’s reasoning settings, , or multimodal processing that could further impact its utility and cost profile.

Related guides & quizzes

Found this useful?