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

A leak says Meta's Watermelon is rolling out soon and is expected to outperform Opus 5.5, GPT-6 Astra and Gemini 4 Argon.

Meta released Muse Wonder (codename Watermelon), raising its Artificial Analysis Intelligence Index score from 31 to 56, behind Claude Opus 5.5 but ahead of GPT-6 Astra and Gemini 4 Argon (53).

A user replied that access to Gemini 4 Argon "just died," suggesting an interruption in a rollout or test channel.

firstadopter joked that Meta's new model shipped before Google's Gemini 4 Argon, reflecting community frustration over Google's repeated delays.

A screenshot shows Meta's Muse Wonder outscoring Gemini Argon and Fable 5.1 on the Artificial Analysis leaderboard, pushing Gemini 4 down the ranking.

A complaint post lists Anthropic Opus 5.5, xAI's grok bot, Google Argon and Meta Muse Wonder as evidence that OpenAI has been overtaken by several labs.

A Polish-language complaint contrasts Gemini 4 Argon going silent after its announcement with Muse Wonder shipping Muse Code right after its own.

A complaint post argues that Gemini 4 Argon will lose its momentum before launch if it keeps limited access and doesn't ship publicly soon.

Comparing Artificial Analysis, DeepSWE, AutomationBench and Terminal-Bench scores plus pricing, Meta's new model leads Argon on most metrics.

A user shared output from Gemini 4 Argon routed through 3.8, saying a public release is rumored this week but the exact day is unknown, with Wednesday a guess that could slip to next month.

A user says Meta's Muse Wonder is a fable-class model that beats GPT-6 Astra, Fable 5.1 and Gemini 4 Argon.

A developer asks jokingly whether Gemini 4 Argon will be released to the public before GTA 6.