Gemini 4 Pro / Flash: benchmarks, hands-on tests and leaks

A post says Gemini 4 Argon scores 53 on AA, 23 points above Gemini 3.1 Pro, ranks first on AutomationBench-AA at 77.5%, has a 15% hallucination rate and a 0.7% prompt injection success rate.

The author expected a minor Gemini 4 update but was surprised by benchmarks pitting Argon against GPT-6 Astra, Claude Fable 5.1 and Opus 5.5, with 13 of 19 categories led outright and strength in coding, finance, legal and long-running agent tasks.

The author says Gemini 4 Argon ranks alongside Astra and Opus 5.5 on intelligence rankings at roughly half the cost, with generous usage limits even on the $20 tier.

A comparison post says Gemini 4 Argon looks as intelligent as GPT-6 Astra and Fable 5.1 but falls behind Opus 5.5, and the author wants to test it soon.

A post says Google launched Gemini 4 Argon, which beats Muse Spark 1.3 and is on par with Astra and Fable 5.1.

Citing Artificial Analysis, a post says Gemini 4 Argon costs only $1.99, below GPT-6 Astra at $3.26 and the Opus 5.5 and Fable 5.1 series at $6-7+.

A short post with no data says Gemini 4 Argon makes GPT-6 Astra, Fable 5.1 and Opus 5.5 look less impressive and that Google is back in the AI race.

A post with a chart claims Gemini 4 outperforms GPT-6 Astra and Opus 5.5 on price-to-performance, framed as a model comparison.

A report says Gemini 4 performs well on industry benchmarks but does less well when employees actually use it for work, raising benchmaxxing concerns.

A post says Argon is already larger than any previous Gemini Pro model and beats Astra and Opus 5.5 in many evals, with a stronger Ultra model still coming.

A post announced as breaking news says Gemini 4 Pro has arrived under the codename Argon and beats existing frontier models in some areas, with benchmark screenshots attached.

The author lists input and output prices for Gemini 4 Argon, Opus 5.5 and GPT-6 Astra, calls Argon notably efficient, and says its multimodal capabilities are worth waiting for.