Introducing Gemini 4 Neon
Google has announced Gemini 4 Neon, a new model designed for fast, multi-step task completion across software engineering, analytical tasks, and scientific research.
2 min read, Wednesday, 18 November

Google has introduced Gemini 4 Neon, a frontier-level model built to handle complex workflows across software engineering, enterprise knowledge work, and scientific research. The model is designed for fast, end-to-end task completion while maintaining high accuracy on demanding professional benchmarks.
Key facts
- Context window
- 1 million tokens
- Input price
- $3 per million tokens
- Cached input price
- $0.15 per million tokens
- Output price
- $15 per million tokens
Capabilities and Benchmarking
Neon ranks #3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its highest thinking level, positioning it just behind larger models. It also ranks #2 on the Vals Index with a score of 70.9 percent and secures the #1 position on both the Legal Research Bench and Tax Agent Bench. In developer evaluations, Neon achieves an Artificial Analysis Coding Agent Index score of 58.6 and records 65.2 percent on Terminal-Bench 4.1.
The model excels at long-running agent workflows, supporting multi-file debugging and refactoring across code repositories. In security tasks, Neon for cyber achieves state-of-the-art performance on SEC-bench Pro with a score of 90.5 percent and reaches 51.26 percent on WebExploitBench v1.2.
Pricing and Context Window
Neon operates with a 1 million-token context window, allowing it to ingest and synthesize large volumes of data from spreadsheets, PDFs, charts, audio, and video simultaneously.
Safety and Alignment
Built on an updated safety stack, Neon maintains strict protections against severe misuse while reducing false refusals. It records low rates of task refusal, declining 1.98 percent of tasks on Terminal-Bench 4.0 and 0 percent on Terminal-Bench Science. The model also demonstrates strong resilience against indirect prompt injection attacks, matching other recent models with a zero percent attack success rate on the Gray Swan benchmark.
Availability
Gemini 4 Neon is rolling out through the Gemini API in Google AI Studio, Google Antigravity for paid subscribers, and the Gemini app for eligible consumer tiers.
Source: Original announcement
Worth a read?



Comments