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

A video pits Gemini 4 against Fable 5.1 using the same prompt, and the creator says the unreleased Gemini 4 wins.

The poster says the Gemini 4 Pro benchmark chart going around is not official but predicts the model lands around that level, with 1M context rather than 2M per the recent leak.

A post lists leaked Gemini 4 Pro specs: ahead of Opus 5.5 and GPT-6 Astra on multiple agentic benchmarks, 2M context, and pricing of $2.25/$11.25 per 1M tokens.

The poster says the benchmark cannot be verified and should be treated as a rumor, doubts the 2M context window but finds the $2.25/$11.25 pricing plausible.

A leak says Gemini 4 Pro may arrive in mid-to-late October with Xhigh and Max reasoning tiers, a 1M context window and 256K output, plus an Argon/barium checkpoint timeline.

A user shares an unverified Gemini 4 Pro leak chart claiming 2M context at $2.25/$11.25, cheaper and higher-scoring than both competitors.

An unconfirmed leak says Gemini 4 may launch in October with xhigh and max reasoning tiers and output raised from 65,536 to 262,144 tokens, keeping the 1M context window.

A post claims leaked Gemini 4 Pro results beat both Anthropic and OpenAI, forcing Fable 5.5 and GPT 6.1 to match its pricing.

A screenshot lists DeepSWE at 88.7% vs 74.2%, pricing at $2.25/$11.25 vs $4/$20, and context at 2M vs 1M for Gemini 4 Pro against Opus 5.5.

A three.js/3D generation comparison video shows Gemini 4 Pro now looking close to Opus 5.5, while Astra takes a completely different approach.

Using one prompt to render the same city, the author says Gemini 4 draws scenes the other model can only describe, and ranks Gemini 4 Pro and Opus 5.5 on top with Astra in between.

A leaked table lists Gemini 4 Pro at DeepSWE 88.7, Terminal-Bench 95.3, OSWorld 86.8, 2M context and $2.25/$11.25 pricing, all above Opus 5.5 and GPT-6 Astra, with the author noting it is unverified.