Gemini 4 Argon leads coding benchmarks at $2/$10 intro pricing

Google's new frontier model tops Opus 5.5 and GPT-6 Astra on DeepSWE, with a 1M-token output ceiling - but wide API access is still gated.

Nowline OCT 2 5:00 PM banner

Top AI stories from the last hour

Top AI stories from the last hour

Copy markdown

  • A new benchmark leader, not just a refresh

    Argon posts 77.9% on DeepSWE v1.1, ahead of Claude Opus 5.5 (74.2%) and GPT-6 Astra (74.1%), ties for first on CWE-bench security remediation (68%), and tops AutomationBench (51.3%). If you gate model switches on evals, this is the one to re-run yours against.

  • $2/$10 now, $4/$20 later

    Introductory pricing is $2 per million input tokens and $10 output, doubling to $4/$20 after the intro window, with a 95% cached-input discount. The output ceiling jumps to 1M tokens (up from 64K), so whole-repo refactors and long-form generation fit in a single call.

  • You probably can't call it yet

    Access is staged - Google AI Ultra subscribers and paid API customers first, with cyber defenders and trusted testers via the Fairwind Program ahead of a wider rollout 'coming soon.' No AI Studio or Vertex GA date yet, so keep shipping on Opus 5.5 or GPT-6 Astra this week.

  • The frontier floor is converging on $2/$10

    Argon lands at the same price as OpenAI's GPT-6.1 Sol (Sep 30), which matches near-Astra quality with 1M context and halved cached input, and Anthropic's Sonnet 5.5 (Sep 28) at $2/$10 with $0.20/M cache reads. Top-tier intelligence is cheap enough now to route by task, not by budget.