Google Launches Gemini 4 Argon with Limited Access and Aggressive Pricing

⚡

Key Takeaways

1Gemini 4 Argon is launched for Pro and Ultra subscribers in the Gemini app and the AI Research mode

2The model was initially tested by "trusted cyber defenders"

3Introductory rates are set at $2 per million input tokens and $10 per million output tokens, with a price increase expected at the end of November

4Google claims superior performance compared to competing models and cites internal use cases

💡Why it matters — This launch marks Google's return to the competition of cutting-edge AI models, with an aggressive pricing offer and claimed capabilities to compete with industry leaders.

Google has launched Gemini 4 Argon, its new generative model, with access restricted to certain subscribers and an aggressive entry price. The launch comes amid a period where its rivals have gained an edge, following a testing phase with cybersecurity specialists.

A Launch in a Catch-Up Sequence Against Meta, OpenAI, and Anthropic

Argon was launched one day after Meta announced its Muse Wonder Full model. This marks Google's first major release since Gemini 3, which dates back to late 2025. In the interim, Anthropic and OpenAI have introduced models deemed impressive, contributing to Google's lagging position. Significant internal upheavals have also raised doubts about the company's ability or willingness to regain a leading position in AI. The arrival of Argon demonstrates that Google remains a player to watch in the AI race, in a context where the balance of power can shift rapidly and competition is not limited to just two companies.

Access Restricted to Pro and Ultra Offers and Pilot Phase Upstream

Gemini 4 Argon is now accessible to Pro and Ultra subscribers in the Gemini app and the AI mode of Search. Ultra subscribers can also access it via the agentic development platform Antigravity. Google has not specified if or when Plus subscribers or free users will be able to use Argon. Prior to this public launch, the model was tested by a select group of "trusted cyber defenders" to assess potential risks.

⚡Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Claimed Capabilities and Initial Internal Uses at Google

Google emphasizes Argon's advanced reasoning capabilities as well as its superior performance in complex tasks, while minimizing hesitations, repetitions, and canned phrases that persist in other AI chatbots. According to the company, Argon outperforms GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 in areas such as coding, cybersecurity, science, finance, and law. Internally, Google uses Argon to identify and correct cybersecurity vulnerabilities, halve the computing power required for YouTube's comment section, and debug the physics simulator for Wing, Alphabet's drone delivery subsidiary. Google claims that Argon is capable of competing with the top-performing models on the market.

Introductory Prices at $2 and $10 per Million Tokens, with Increase at the End of November

The model is launched with an introductory price of $2 per million input tokens and $10 per million output tokens, which represents about one-fifth of the cost of OpenAI's GPT-6 Astra and Anthropic's Fable 5 models. Google plans to double these rates at the end of November, following the introductory period. Even after this increase, Argon will remain significantly cheaper than competing top models on a per-token basis. However, Google clarifies that lower per-token rates do not necessarily mean lower overall costs, as each model may consume a different number of tokens for the same task.