Qianwen Qwen3.8 official announcement! 2.4T parameters are open source, pointing directly to Fable 5

Zhidongxi Author | Wang Han Editor | Xinyuan Zhidongzhi reported on July 19 that just now, Alibaba's Qianwen large model officially announced that Qwen3.8 will soon be released and open source, with new model parameters of 2.4T. Qianwen calls this model “probably the most powerful model besides Fable 5.” At present, the Qwen3.8-Max preview version model has been launched on Alibaba Token Plan, Qoder and QoderWork. The new model has a limited-time discount, with a 10% discount on credit consumption during the day and even lower at night. Both the individual and team versions of Token Plan have discounts, starting from as low as 35 yuan per month. Experience address: Qianwen AI Platform: https://platform.qianwenai.com/pricing/token-plan In the past half month, domestic large models have started blowout mode again. This week, Kimi-K3 was launched as open source with a total parameter of 2.8T. Its long-range tasks, front-end rendering, Agent Swarm and other capabilities have won unanimous praise from developers at home and abroad. Musk even commented "impressive". Russ Salakhutdinov, the doctoral supervisor of Kimi founder and CEO Yang Zhilin and a leader in machine learning at CMU, also posted a message praising the new version of Kimi for bringing major breakthroughs to the open source community. In addition, according to overseas developers, the official version of DeepSeek-V4 has been gray tested for some users and is expected to be officially released tomorrow. According to The Information, MiniMax’s new generation large language model M3 Pro will also be released soon, with a parameter scale of 2.5 trillion to 3 trillion, and plans to open source it.