Progression

2 AI models, Qwen-3.8-Max-Preview and Kimi-K3 are going into a head-to-head competition with the best publicly available frontier AI models from USA right now, aka, Fable 5 and GPT-5.6 Sol, while being open weight models.

Yes, there are many instances where they are behind these 2 (Fable & Sol) by a considerably good margin, but there are also many instances where they are actually out-accelerating and leading.

It's not just benchmarks, they DO have a big model smell and big model taste. Qwen-3.8-Max-Preview and Kimi-K3 are literally 2.4 Trillion and 3 Trillion parameter open weight AI models from China directly competing for the frontier of AI with USA.

READ THAT SENTENCE AGAIN!!! THIS IS JULY 2026!!!!

Xi Jinping, at the World AI conference, held in 2026 just very recently, has clarified in clear cut words that China is hellbent on accelerating the frontier of AI models through open weights, open source and open collaboration across multiple countries and unions of countries like the African Union, ASEAN (Association of South East Asian Nations), BRICS (Brazil, Russia, India, China, South Africa), the Arab Union etc etc etc

Yes, there's no doubt in the fact that except for SWE & cybersecurity tasks, the State-of-the-Art US AI models like Claude Fable 5 and GPT-5.6 Sol have a much more comfortable lead across all other domains of expertise, but if you have any intuition and imagination developed from experiencing the past 3 years of AI progress, you just know that "IT Will All Fall To SCALE"

There's also no doubt in the fact that the Frontier US AI labs are making increasingly accelerating strides through Proto RSI since December 2025 and have been training drastically better AI models over the past 3 months already. In fact, the leap is so big that OpenAI prefers to call it GPT-6 straight from GPT-5.6 Sol itself, and every single GPT-5.x iteration has been a total banger, so just imagine.......

but hey, here's the thing !!!

The frontier open weight Chinese labs have bulldozed through every single expectation wall based on expected limitations like compute, cost, data etc, etc, etc, , be it in pre-training or post-training , from being one year behind, to 7 months behind, to 5 months behind, to co-runners... and we're only just getting started!!!!

Imagine Dario and Anthropic after doing all this fear mongering for months lol.It's all good and it just turned out to be all right......

A few years ago, during the GPT-4 days or earlier, Sam Altman, during a meetup with students from India, claimed that there was no way of competing with them. Look at the situation only 4-5 years later lol lmao XD

By the way, inserted: the first 1.5 Trillion AI model from India is in the process of training right now, and India has also become the 3rd country after USA and China to have a private startup that can launch rockets in the low orbit of earth with a fully 3-D printed engine. Oh and also, ex-OpenAI researcher Shyamal Anadkat, also returned to India earlier this year and has been silently cooking with fellow researchers and investors with the goals of building superintelligence.

All-in-all, this is the best possible timeline for open, equitable, & frontier AI access for everybody across the globe. Historic moment (July 2026)!!!!

The greatest possible outcome in the most likely....cuz why not 😎❤️‍🔥