In the "life-or-death race" of commercializing large models,
On March 30, according to
Computing power quotas are hard to come by: millions paid in advance to secure supply
With the breakthrough performance of
Supply tightness: According to insiders, after the launch of the K2.5 model, the API supply's TPM (tokens per minute) quota was quickly exhausted.
Spending millions: To gain an advantage in the computing power competition, some major customers have made commitments worth tens of millions of dollars, and even offered prepayment guarantees to obtain
Technical foundation: From code prodigy to "Agentic" pioneer
Beneath the explosive revenue of
Model iteration: In July last year, the company released and open-sourced a trillion-parameter model
Global top ranking: After the release of the upgraded version
Industry observation: Large models have entered the revenue era where users vote with their feet
Previously, the industry focused more on model parameters and technical benchmarks, but now the focus is rapidly shifting towards "revenue-generating ability":
Revenue cycle:
Capital anchor point: In the capital market, revenue scale has become a key indicator to measure whether a startup in large models can move from the "laboratory" to the "industrial center."
Conclusion: $100 million is just the beginning
From the open-source exploration of


