Physix Frontier · News Briefing Card (GeekPark · Nov 17, 2026)

Alibaba launches Qwen4-Max, cutting per-task cost by 58%

KEY FACTS

  • Alibaba Qwen officially released its flagship model Qwen4-Max on November 16, with 2.4 trillion total parameters.
  • The model natively supports text, image, and video input, with a context window of 1 million tokens.
  • In real-world work scenario evaluations, the average cost per task dropped from 26.58 yuan to 11.16 yuan.
  • Only 71B parameters are activated per token, roughly 25% fewer than the previous generation.
  • It scored 52.69 on the Work360 industry benchmark, up from 47.74 for its predecessor.

KEY DATA

2.4 trillionTotal backbone parameters
1 million tokensContext window
11.16 yuanAverage cost per task
52.69 pointsWork360 score

PHYSIX OBSERVATION

The competitive focus of flagship models is shifting from benchmark scores to the cost ledger of real workflows. Qwen4-Max slashes per-task cost by nearly 60%, and combined with low cache pricing, it aims squarely at high-frequency Agent invocation scenarios. The signal to enterprises is clear: when choosing a model, the bill is more persuasive than the leaderboard, and only models that make it onto the shift schedule count as a win.

Source: GeekPark report