Gemini 4 Argon Ties GPT-6 Astra on Index at Lower Launch Cost

On the Artificial Analysis Intelligence Index, Google's Gemini 4 Argon has tied OpenAI's GPT-6 Astra, both scoring 53 points. The per-test-task expense is also lower given the current launch pricing.

Google DeepMind unveiled the model on September 30. Access is limited initially to a set of trusted cybersecurity specialists.

Introducing Gemini 4 Argon – our new frontier model.

It's built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program. pic.twitter.com/X8acOWJOSF

— Google DeepMind (@GoogleDeepMind) September 30, 2026

Argon Returns Google to Top 3 AI Labs

The independent benchmarkers at Artificial Analysis stated that the result places Google again in the top three AI laboratories. Argon also beats GPT-6.1 Sol by one point; Sol managed a score of 52.

Argon outperformed the earlier non-Flash model, Gemini 3.1 Pro Preview, by 23 points. Earlier in September, Google had released Gemini 3.8 Flash alongside a cybersecurity variant. SpaceXAI's Grok 4.7, by comparison, achieved a score of 46 on the same index in September.

Artificial Analysis credited the improvements to a lower hallucination rate and better agentic capabilities. In the AA-Omniscience test, Argon had a 15% hallucination rate, compared with 51% for Astra. However, Argon got 50% of the questions correct, 13 points below Astra.

Argon topped the AutomationBench-AA benchmark with 77.5%, outperforming Anthropic's Claude Sonnet 5.5, which posted 71.3%. However, it lagged behind Sonnet 5.5, Claude Opus 5.5, and Astra on Terminal Bench 4.

Introductory Pricing Lowers the Bill

Google established launch prices of $2 per million input tokens and $10 per million output tokens. After the promotional period, those figures will double to $4 and $20.

With the discount, Artificial Analysis calculated Argon's per-task cost at $1.99, while Astra's stood at $3.26. At regular prices, the cost climbs to $3.98, roughly 1.2 times Astra's.

The savings stem from lower per-token prices, even though Argon consumes more tokens. On average, Argon used 62,000 output tokens per task, versus Astra's 27,000. The model also supports a 1-million-token output limit.

“To support Gemini 4 Argon’s capabilities across longer, more complex use cases, we are significantly expanding the model’s output token limit to an industry-leading 1M tokens, up from the previous 64K tokens,” the team said.

Cyber Defenders Get First Glimpse of Gemini 4 Argon

Google CEO Sundar Pichai noted that the model is already in heavy use by Google teams, for tasks spanning coding through quantum computing.

“Importantly, Argon has frontier safeguards, and we are rolling it out responsibly – it’s with the US gov’t and going to a set of trusted cyber defenders through our Fairwind Program today,” he noted.

According to Google's launch announcement, Argon can independently find, validate, and patch serious software vulnerabilities. As a result, trusted defenders and internal Google teams will receive it without cyber guardrails.

When the rollout broadens, paying API users and Google AI Ultra subscribers will be first to get access. Google has not yet revealed if the introductory pricing will still be in place at that time.

Google also stated that it is taking part in the US government's voluntary pre-release access process.