[AINews] Gemini 4 Argon: GDM’s answer to Astra/Fable, with 1M output

Google DeepMind introduced Gemini 4 Argon, a model aimed at coding, enterprise knowledge work and cyber defense, according to posts from the company and CEO Sundar Pichai. Argon is Google's first larger-than-Flash release since Gemini 3.1 Pro shipped in February, following incremental 3.x Flash versions and a management shakeup last month.

Availability is limited at launch. Access begins with government users and trusted cyber defenders in the Fairwind Program, and Google says it will refine guardrails before opening access to developers, enterprises and consumers. Google describes access as coming "as soon as possible."

Argon carries an industry-leading 1M-token output limit, up from 64K, per Google. Vals lists 262K max output; Artificial Analysis says it reached 1M output tokens through Long Decode Continuation, a new API feature that pauses long responses and resumes them across calls. Pricing is $4/$20 per 1M input/output tokens, with a 50% introductory discount to $2/$10 and no announced end date. Cached input gets a 95% discount.

Google claims Argon takes first place on 13 of 19 published benchmarks against GPT-6 Astra and Claude Opus 5.5. It reports 77.9% on DeepSWE versus 74.2% for Opus 5.5 and 74.1% for Astra, that Argon agents freed more than 300 TiB of data-center memory, and that they are migrating more than 800K lines of C/C++ kernel code to Rust. The team also says internal agent loops built on Argon helped complete the CK conjecture.

Independent evaluations are mixed. Artificial Analysis scores Argon 53 on its Intelligence Index, matching GPT-6 Astra and edging GPT-6.1 Sol at 52, at $1.99 per task discounted versus $3.26 for Astra, with Argon averaging 62K output tokens per task against Astra's 27K. Vals ranks Argon first on its index at 68.9% and reports 57.6% on Terminal-Bench 4.0, up from 19.0%. Observers questioned some published numbers, including a reported 19.6% on Harvey's legal benchmark and the DeepSWE figure.