Alibaba launches 2.4T-parameter Qwen3.8, claims it rivals top AI models

Alibaba ↘ 1.17% announced on Saturday the launch of Qwen3.8, a 2.4 trillion-parameter large language model that the company says is "second only to Fable 5" in overall performance. A preview version, Qwen3.8-Max-Preview, is already available to developers through Alibaba's Token Plan subscriptions and its coding platforms Qoder and QoderWork, with open weights expected to follow.digg+2

A New Scale for Qwen

The model represents a major jump in scale for Alibaba's Qwen series, which released a 1 trillion-parameter Qwen3-Max in September 2025 and a 397-billion-parameter open-weight Qwen3.5 in February 2026. Qwen3.8 more than doubles the parameter count of its predecessor and is described as fully multimodal.linkedin+4

Alibaba's Qwen team announced the model on X, stating it would go open-weight "soon" — a departure from the company's more recent practice of keeping flagship Max models proprietary and API-only. The Qwen3.7-Max model released in May 2026, for instance, was never made available as open weights.yottalabs+3

Performance Claims Lack Independent Verification

Alibaba claims Qwen3.8's overall performance ranks just below Anthropic's Fable 5, but the assertion currently rests entirely on internal evaluations. No independent third-party benchmarks have been published. This mirrors a pattern from previous Qwen releases: when Alibaba launched Qwen3.5 in February, it reported benchmark results comparable to GPT-5.2 and Claude 4.5 Opus, though outlets including CNBC noted those comparisons were self-reported and could not be independently verified.cnbc+2

The AI model race between Chinese and American labs has intensified throughout 2026, with Alibaba releasing a succession of models — Qwen3.5, Qwen3.5-Max, Qwen3.7-Max, Qwen3-Max-Thinking — each claiming competitiveness with frontier systems from OpenAI, Anthropic, and Google DeepMind.reuters+2

Access and What Comes Next

Developers in mainland China can access Qwen3.8-Max-Preview now through Alibaba's Token Plan subscription tier. The promise of open weights, if fulfilled, would mark a return to the strategy that made earlier Qwen models among the most widely adopted open-source AI systems globally. Alibaba has not announced a timeline for the full release or provided details on architecture, training data, or context window specifications.kucoin+2