Today Tencent Hy3 official version released

WorkBuddy follows up simultaneously
Now in WorkBuddy, Hy3 is free for a limited two-week period
WorkBuddy is similar to CodeX
Workflow assistant, real workflows → fuel for model iteration
This is the fourth layer of scaling law
The expanded target shifts from parameters to task feedback
Hy3 just launched today:
• Enters the national model T1 (alongside K2.6 / Qwen3.7 / DeepSeek V4)
• 295B total parameters, 21B activated, the smallest and cheapest in T1
• Input 1 yuan, Output 4 yuan, Cache 0.25 yuan
• GLM 5.2 is still the strongest, but 7x more expensive, 2x more resource-consuming
The next battle in AI is at the "workflow entry point"
Domestically, WorkBuddy is the leader, with users equal to the sum of the 2nd to 4th rankings
So the logic is:
Base model enters T1 (Hy3, verified today)
→ Secures the largest scenario (WorkBuddy, simultaneous)
→ Data flywheel spins up
→ Narrative shifts from "catching up" to "pricable"
For the stock price:
If the market re-incorporates the "AI workflow entry point" into the valuation anchor
This is the expectation gap
It is not currently priced in
But today's action is a signal
Limited two-week free trial
Window period, see for yourself