Gemini 4 Argon is a new AI model designed to excel across enterprise workflows, with strong reasoning, coding performance, and the ability to sustain long, multi-step tasks. It also offers unmatched multimodal understanding, allowing it to handle a wide range of complex business needs.
In benchmark testing, Argon posts strong results in knowledge work. It scored 68.9% on the Vals Index and 51.3% on Automation Bench, and it leads in agentic coding with 77.9% on Deep SWE v1.1. On Harvey's Legal Agent Benchmark it reached 19.6%, well ahead of competing models. It also leads in long-context tasks, with 99.7% on Graph Walks up to 128k tokens.
Argon is also positioned as a leader in defensive cybersecurity. The model can autonomously find, validate, and patch critical software vulnerabilities, and it scored 68.0% on CWE-bench v1. Google says it is designed to better equip defenders for a new era of cyberattacks.
On safety, Google says Argon is built to refuse harmful requests while still supporting legitimate, dual-use scientific research under its Frontier Safety Framework. The company says it is its most resilient model yet against indirect prompt injection attacks, and it is deploying misalignment mitigations that monitor the model's chain-of-thought and actions and halt execution when necessary. Sandboxed environments are also being isolated and sealed before high-risk training or evaluations begin.
Source: Hacker News · Summarized by HeadlinesBriefing