Test Driving Alibaba's Frontier Model: Qwen 3.8 Max
I am looking for a new daily driver model
I am looking for a new daily driver model, or at least one that can offset Anthropic’s ridiculous pricing and OpenAI’s almost as ridiculous pricing. Today I am testing Alibaba’s frontier model - Qwen 3.8 Max preview - which runs less than half the cost of Anthropic’s Opus model family.
I am running Qwen 3.8 Max inside of Qwen Code (it is a 2.4 trillion parameter model, too large for me to run locally), so far so good.
I am running Qwen 3.8 Max on a personal project as I don’t want to risk it breaking Kilby or TEK2dayFIN. I will report results back here.
Later this week Alibaba is to release Qwen 3.8 27B, a smaller frontier-level model which I should be able to run locally at zero cost (other than hardware costs).
If I have success with Qwen 3.8 27B as a daily coding model, my plan will be to adapt a version of it - a TEK2day-labeled model - that I will optimize for Finance/Investment Management and house inside of Kilby. I will also make that model available to everyone - either opensource on Github or for a small fee, I have yet to decide and won’t know until I dig in.
Learn more about Qwen 3.8 Max HERE.
