Archive · 1 Oct 2026
Google Announces Gemini 4 Flagship Argon After Months Of Delays
Google announced a powerful new AI model called Argon.
It will be part of the company’s Gemini 4 family.
Google says Argon is bigger and better for difficult tasks than its earlier Pro models.
The company says it performed as well as some leading models from OpenAI and Anthropic in certain tests.
Argon did better in some tests but did worse in two coding tests.
Google is giving the model to selected cybersecurity partners for now.
The company has not said when everyone will be able to use it.
Google also decided not to release the previously planned Gemini 3.5 Pro model.
Google announced Argon, a new top-tier AI model intended to anchor its Gemini 4 generation.
The company says Argon is larger than previous advanced Pro models and performs strongly on complex workloads.
Google said Argon compares with OpenAI’s Astra and Anthropic’s Opus on selected coding and cybersecurity benchmarks.
Company-reported results showed Argon ahead on some benchmarks but behind on two of four coding benchmarks cited.
Google canceled plans for Gemini 3.5 Pro and has not announced when Argon will become publicly available.
- Who
- Google announced the Argon model, while OpenAI and Anthropic were cited as competitors.
- What
- Google introduced a flagship model for its Gemini 4 generation and canceled plans for Gemini 3.5 Pro.
- When
- The announcement was made on Wednesday; the articles do not provide a specific date.
- Where
- The announcement was reported from San Francisco.
- Why
- Google said Argon is designed for complex workloads and to help it compete with leading AI models from Anthropic and OpenAI.
Google’s Position
Evidence Of Remaining Gaps
Overall capability
Google’s Position
Google described Argon as its most performant model and said it was comparable to leading models on key coding and cybersecurity benchmarks.
Evidence Of Remaining Gaps
The company-reported results showed Argon remained behind Astra and Opus on some metrics, including two of four coding benchmarks.
Position in the AI race
Google’s Position
Google has emphasized Argon’s capabilities and previously rejected the idea that it was losing ground.
Evidence Of Remaining Gaps
The months of delays, the canceled Gemini 3.5 Pro release, and continued releases from Anthropic and OpenAI have been cited as signs that Google fell behind.
Model availability
Google’s Position
Google is making Argon available to selected cybersecurity partners and participating in a voluntary pre-release access process.
Evidence Of Remaining Gaps
The model does not yet have a public release timeline, limiting immediate access for general users.
- Model
- Argon
- Model family
- Gemini 4
- Previous planned model
- Gemini 3.5 Pro, which Google no longer plans to release
- Public availability
- No release timeline was provided
- Initial access
- Select cybersecurity partners
- Benchmark results
- Argon led on some company-reported benchmarks but trailed on two of four coding benchmarks cited
- Competitors cited
- OpenAI’s Astra and Anthropic’s Opus