Google releases Gemini 4 Argon globally, positioning it against GPT-6 Astra and Claude Fable 5.1

Google's Argon launch and tiered access

Google has rolled out Gemini 4 Argon globally, making the new flagship model available to developers, businesses and consumers beginning Oct. 13. Access is tiered: AI Ultra subscribers can use Argon on the agentic development platform Antigravity, while AI Ultra and Pro subscribers get access in the Gemini app and Google Search's AI mode. Free Gemini users and paying Plus subscribers are limited to the weaker Flash models under Google's revised tier structure, with no timeline confirmed for broader availability [The Verge; The Decoder].

Argon replaces the Gemini Pro class as Google's new frontier series, and Google says additional Gemini 4 models will follow. The Decoder reports Google is internally testing a potential successor codenamed "Carbon," described as targeting coding performance in particular [The Decoder].

Pricing and benchmark positioning

Google is introducing Argon at $2 per million input tokens and $10 per million output tokens during a promotional period, with a stated 95 percent discount on cached inputs. The company says the rate is roughly a fifth of the cost of OpenAI's GPT-6 Astra and Anthropic's Claude Fable 5 models, and that Argon will remain significantly cheaper per token even when the promotional price roughly doubles at the end of November [The Verge].

On Google's own benchmarks, Argon is claimed to outperform GPT-6 Astra, Claude Fable 5.1 and Claude Opus 5.5 across coding, cybersecurity, science, finance and law. In independent rankings reported by The Decoder, Argon holds the top spot on the Vals Index but still trails Anthropic's Claude Opus 5.5 in overall rankings, and sits behind Meta's Muse Wonder Full, which launched a day earlier [The Decoder].

Internal use cases and self-reported engineering gains

Google says Argon is already deployed internally. Reported uses include finding and patching cybersecurity vulnerabilities, halving the computing power required for YouTube's comments section, and debugging the drone physics simulator for Wing, Alphabet's drone delivery subsidiary [The Verge]. The company also shared self-reported engineering metrics: software engineers using Argon increased agent-driven code submissions by more than 35 percent while code rollbacks dropped by nearly a third [The Decoder].

In the Chromium project, Argon helped the Chrome team complete an architectural refactor across hundreds of files within a 40-million-line codebase, reducing work that Google said would have taken months to a few weeks. Working with Google's open-source Agent Development Kit, Argon ran autonomously overnight and produced optimizations that cut session preparation time by 40 percent and sped up cold imports by 21 percent. Inference speed for the open Gemma models on an A100 via vLLM rose from roughly 37,900 to over 108,000 tokens per second, a factor of 2.85 [The Decoder].

Reported capabilities and demos

Google describes Argon as delivering "frontier reasoning and improved performance across complex workflows" while reducing hedging, repetition and clichés in chatbot outputs [The Verge]. The Decoder reports the model can generate complex 3D models, games and animations in pure code that run directly in the browser, including a raytracing engine written entirely in Python without external libraries. In a separate demo, Argon started from a single image in Antigravity and controlled Blender via MCP in an autonomous loop, modeling gothic architecture and setting up audio-reactive lighting to produce a finished 3D scene [The Decoder].

Next milestone

Google's promotional pricing for Argon is scheduled to roughly double at the end of November, returning to $4 per million input tokens and $20 per million output tokens according to the price table published alongside the launch [The Decoder; The Verge]. Additional Gemini 4 variants are expected to follow the Argon release.