🚨 Gemini 4 Pro : Specs Leaked and they're insane 🔥
Google’s next flagship might be MUCH more powerful than expected. 👀
>2M token context window
>256K output limit reportedly being tested
>$2.25 / 1M input tokens
>$11.25 / 1M output tokens
>95.3% on Terminal-Bench 2.1 in the leaked table
>88.7% on DeepSWE v1.1
>86.8% on OSWorld-2.0
>94.7% on CharXiv Reasoning
>Reportedly ahead of GPT-6 Astra on several leaked benchmarks
And the wildest part?
The leaked table puts Gemini 4 Pro at $2.25/M input tokens potentially making it dramatically cheaper than the other frontier models listed.
Google has already confirmed that Gemini 4 is in its most ambitious pre-training run yet. Now alleged internal checkpoints, benchmark tables and Arena sightings are starting to surface.
If these numbers are real, Google is cooking 🔥👀
These specs, prices and benchmark results are from leaks and have NOT been officially confirmed by Google.
Is Tibo even aware of his own product???
Tibo asking “Sol supports 1M, tell me more?” while Codex users are still dealing with ~272k context limits is kinda wild.