According to @DeepLearningAI, Alibaba released Qwen3-Next-80B-A3B in Base, Instruct, and Thinking variants under an open-weights Apache 2.0 license, targeting faster long-context inference and supporting inputs up to 262,144 tokens with multi-token prediction; source: DeepLearning.AI on X, Sep 22, 2025, https://twitter.com/DeepLearningAI/status/1970254860416131146; The Batch overview, https://hubs.la/Q03KsR8W0. The 80-billion-parameter mixture-of-experts replaces most vanilla attention layers with Gated DeltaNet and the remainder with gated attention, is trained on a 15-trillion-token subset of the Qwen3 dataset, and is fine-tuned with GSPO; source: DeepLearning.AI on X, Sep 22, 2025, https://twitter.com/DeepLearningAI/status/1970254860416131146; The Batch overview, https://hubs.la/Q03KsR8W0. For trading focus, key measurable specs to track are the 262,144-token context window, multi-token prediction, and open-weights Apache 2.0 licensing, as these parameters define model accessibility and performance for builders; the source does not mention any cryptocurrency integrations or market pricing effects; source: DeepLearning.AI on X, Sep 22, 2025, https://twitter.com/DeepLearningAI/status/1970254860416131146; The Batch overview, https://hubs.la/Q03KsR8W0.