Google's AI is now substantially behind, even losing to freely available Chinese Open Weight Models. The market hasn't caught on yet

Moonshot AI just launched Kimi K3, an open weight model that beats OpenAI's GPT 5.5 and Anthropic's Opus 4.8, but trails 5.6 Sol/Fable 5.

Notably, Google's best performing model, 3.5 Flash scores 37% on DeepSWE, compared to 67.5% for Kimi K3. They also lose to GLM 5.2(44%), Meta's Muse Spark(53%), and SpaceX's Grok 4.5(54%).

There is absolutely no reason to pay for Gemini's API when you can just host GLM 3 yourself or use a cheaper competitor like Grok 4.5/Muse Spark/GPT 5.6 Luna, which perform better for less cost.

The rumor is Google had to delay 3.5 Pro because it had serious issues and did not meaningfully improve performance over 3.5 Flash.

It's insane that despite theoretically being in one of the strongest positions(tons of access to user data/web data, huge amounts of compute, huge amounts of capital), they're losing to Chinese labs that have to deal with export restrictions and limited compute.

The reality is that Google has a serious executive leadership problem that prevents them from competing with other AI providers. Their CEO has no background in tech(he has an MBA), and makes poor decisions. Good opportunity to sell if you have any shares.