Google delays Gemini 3.5 Pro after coding stumbles as China's Kimi K3 tops the leaderboard

Google's next flagship artificial-intelligence model has hit a speed bump. The launch of Gemini 3.5 Pro was reportedly delayed after the model fell short on coding and complex-reasoning tasks in internal testing, and Alphabet shares fell about 4% on the news.

The stumble is a reminder that progress at the frontier of AI is no longer guaranteed to be smooth or linear, even for the best-resourced labs in the world.

The gap has not gone unfilled. Moonshot AI's Kimi K3, a 2.8-trillion-parameter open model from China, took the top spot on a major coding leaderboard — a striking marker of how quickly open and non-US developers are closing in on, and sometimes overtaking, the established leaders.

The competitive picture is broad. Reporting this month indicated that OpenAI released its most advanced model family yet after clearing a US government review, while Meta green-lit production of its own custom AI chip to reduce its dependence on outside suppliers.

Regulators are shaping the field as aggressively as the engineers. The European Commission adopted binding requirements under the Digital Markets Act ordering Google to open Android to rival AI assistants and to share portions of its search data with competitors, including AI developers.

That combination — a delayed flagship, a surging open-model rival and a forced opening of a dominant platform — captures the central tension of AI in 2026: raw capability is advancing fast, but the ground rules are being rewritten just as quickly.

For developers and businesses, the practical takeaway is choice. A market where several capable models compete, and where regulators push for interoperability, tends to mean lower prices and less lock-in than one dominated by a single provider.

Google, for its part, will be under pressure to ship a Gemini update that clearly reclaims the lead — and to do so before rivals cement the perception that the frontier has slipped from its grasp.