Gemini 3 Pro Arrives: Hailed as New AI Leader Amidst 'Unprecedented Capability Jump'

Google has launched Gemini 3 Pro, a new iteration of its large language model, which is being described by early users and evaluators as a ‘massive’ release signifying an ‘unprecedented capability jump’ on par with the impact of GPT-4. The model demonstrates exceptional prowess in design, UI generation, and multimodal understanding, processing audio, video, images, and documents like PDFs seamlessly. Benchmark results underscore this leap: Gemini 3 Pro achieved 45.8% on Humanity’s Last Exam (with tools), 92% on GBQA Diamond, and a significant 31.11% on ARC AGI 2, marking a 2x state-of-the-art improvement. Its three-dimensional reasoning capabilities, notably highlighted in Minecraft benchmarks, appear to be currently unrivaled. Artificial Analysis’s intelligence index has positioned Gemini 3 Pro as the new leader, outscoring GPT 5.1 by three points.

Despite its groundbreaking intelligence, Gemini 3 Pro presents a nuanced user experience. While lauded for its ‘one-shotting’ ability, improved writing quality, and enhanced tool calling over previous Gemini versions, it exhibits a high hallucination rate of 88% on Artificial Analysis’s Omniscience Index, on par with Gemini 2.5 Pro. Users report that the model can be ‘quirky,’ occasionally ignores instructions, gets stuck in loops, and requires active ‘babysitting’ for verification—a sentiment echoed by industry figures like Matt Schumer. Furthermore, while showing improved token efficiency over Gemini 2.5 Pro, Gemini 3 Pro’s token usage is still higher than some competitors, combined with a higher per-token cost ($2 in / $12 out per million tokens up to 200k context), resulting in significantly more expensive benchmark runs. Developer experiences also note occasional CLI hangs and editor integration challenges. Despite these operational eccentricities, the consensus points to a powerful model that, when properly managed, delivers ‘brilliantly, almost humanly right’ results, solidifying Google’s position as a dominant, albeit ‘quirky,’ force in the rapidly evolving AI landscape.