Gemini 4 Argon (High): Intelligence, Performance, and Price Breakdown
benchmarks gemini google
| Source: HN | Original article
An analysis examines Google's Gemini 4 Argon (High) on intelligence, performance and price, benchmarking it against other AI models.
Google’s DeepMind has published a detailed analysis of its newest flagship model, Gemini 4 Argon (High), laying out how the system stacks up on intelligence, speed and cost. The report shows the model scoring 53 on Artificial Analysis’s Intelligence Index – a rating that puts it on par with the newly announced GPT‑6 Astra and Claude Fable 5.1, and just behind Claude Opus 5.5 and Sonnet 5.5. Performance metrics also highlight a dramatic jump in context length: Argon now supports up to one million output tokens, a ten‑fold increase over the 64 K limit of earlier Gemini versions. Speed figures released in the benchmark tables indicate a strong tokens‑per‑second rate and a reduced time‑to‑first‑token, which the analysis credits to architectural refinements aimed at “complex workflows across real‑world software engineering, enterprise knowledge work and cybersecurity defense”【Gemini 4 Argon (High) Intelligence, Performance & Price Analysis】.
Pricing is positioned as a market‑disruptor. At launch, Google charges $2 per million input tokens and $10 per million output tokens – a 50 % discount compared with the rates announced for the model’s predecessor, and a figure that is expected to rise to $4/$20 later【Gemini 4 Argon (High) Intelligence, Performance & Price Analysis】. The low entry cost, combined with the expanded token window, makes Argon attractive for high‑volume, context‑heavy applications such as legal research, financial analysis and threat hunting.
Why it matters is twofold. First, the model’s intelligence rating signals that Google is back in the “frontier” tier of AI, directly challenging the latest offerings from OpenAI and Anthropic. Second, the pricing strategy could pressure rivals that rely on higher‑margin API fees, especially as Anthropic’s agentic AI now accounts for the bulk of its revenue【2026-10-01, id 14772】.
What to watch next includes the rollout of broader access beyond the current “trusted cyber‑defenders” beta, the scheduled price increase, and how enterprise adopters respond to the combination of massive context windows and competitive pricing. The next few months will reveal whether Argon can translate its benchmark lead into real‑world market share.
Sources
Back to AIPULSEN