Gemini 4 Argon

# Gemini Source: [https://deepmind.google/models/gemini/](https://deepmind.google/models/gemini/) - [Capabilities](https://deepmind.google/models/gemini/#capabilities) - [Performance](https://deepmind.google/models/gemini/#performance) - [Safety](https://deepmind.google/models/gemini/#safety) - [Ecosystem](https://deepmind.google/models/gemini/#ecosystem) ### Excels in coding & enterprise knowledge work Argon’s reasoning capabilities across subjects, coding performance, ability to sustain long, multi\-step tasks, and unmatched multimodal understanding enable it to excel across a range of enterprise workflows\. ### Leading in defensive cybersecurity The model can autonomously find, validate, and patch critical software vulnerabilities\. BenchmarkNotesGemini 4 ArgonGPT\-6 AstraClaude Fable 5\.1Claude Opus 5\.5Vals IndexKnowledge work**68\.9%**63\.1%65\.8%67\.0%AutomationBenchKnowledge workScore**51\.3%**41\.4%31\.4%42\.5%Vals Finance Agent v2Knowledge work**65\.4%**53\.5%58\.9%58\.6%Harvey's Legal Agent BenchmarkKnowledge work**19\.6%**5\.4%6\.7%3\.8%DeepSWE v1\.1Agentic coding**77\.9%**74\.1%67\.4%74\.2%FrontierSWE v2Agentic coding55\.0%**65\.5%**56\.3%62\.3%Vibe Code BenchAgentic coding**91\.9%**89\.6%90\.3%90\.3%Terminal\-bench 4\.0Agentic coding57\.4%58\.2%57\.9%**66\.4%**PostTrainBenchML engineering45\.3%44\.3%40\.2%**49\.3%**Terminal\-Bench Science 0\.1Science and math57\.6%**68\.1%**52\.6%63\.3%LABBench 2Science and math**88\.8%**85\.4%68\.6%73\.1%RiemannBenchScience and math**76\.0%**72\.0%65\.6%69\.6%GraphWalksLong contextUp to 128k, BFS \(F1\)**99\.7%**98\.7%91\.4%90\.6%256k to 1M, BFS \(F1\)**84\.2%**71\.8%65\.0%66\.8%Agent's Last ExamComputer usePass rate**39\.5%**34\.2%—38\.2%OSWorld\-2\.0Computer useOffline subset Partial score69\.2%**72\.6%**——ChartographyMultimodal understanding**71\.6%**71\.0%46\.2%66\.3%LVBenchMultimodal understanding**91\.7%**87\.5%79\.7%83\.7%CWE\-bench v1Cybersecurity**68\.0%****68\.0%**58\.0%67\.0% Slide 1 of 4 ### Defending against misuse Argon is designed to refuse harmful requests while preserving legitimate, dual\-use scientific research, as per our Frontier Safety Framework\. ### Defending against prompt injection attacks Argon is our most resilient model yet against indirect prompt injections\. ### Monitoring for misalignment We are deploying misalignment mitigations that monitor Argon’s chain\-of\-thought and actions and stop execution when necessary\. ### Hardening systems We are hardening our sandboxed environments by isolating and sealing them before high\-risk training or evaluations begin\. ### Defending against misuse Argon is designed to refuse harmful requests while preserving legitimate, dual\-use scientific research, as per our Frontier Safety Framework\. ### Defending against prompt injection attacks Argon is our most resilient model yet against indirect prompt injections\. ### Monitoring for misalignment We are deploying misalignment mitigations that monitor Argon’s chain\-of\-thought and actions and stop execution when necessary\. ### Hardening systems We are hardening our sandboxed environments by isolating and sealing them before high\-risk training or evaluations begin\. Gemini 4 Argon with cybersecurity defense capabilities ## To better equip defenders for the new era of cyberattacks, we trained Argon to be highly capable at cybersecurity defense Slide 1 of 4 ### Gemini Omni Create anything from anything, starting with video ### Nano Banana Create and edit images with Gemini ### Gemini Audio Push new frontiers with intuitive inputs, natural expressiveness, and the ability to take action ### Gemini Robotics The intelligence layer to power any kind of robot ### Gemini Omni Create anything from anything, starting with video ### Nano Banana Create and edit images with Gemini ### Gemini Audio Push new frontiers with intuitive inputs, natural expressiveness, and the ability to take action ### Gemini Robotics The intelligence layer to power any kind of robot