Alibaba ships Qwen 3.7 with 2M-token context as default — extends multilingual leadership across 200+ languages and undercuts Western pricing
Alibaba's Qwen team rolled out Qwen 3.7 with a 2-million-token default context window, up from the 262K context of Qwen3-Max with 1M extension. The vision-language model now supports 201 languages, makes the largest open-source context-window claim, and lists at roughly $1.20 input / $6 output per million tokens — substantially below GPT-5.5 and Claude Sonnet 4.6 for comparable quality tiers.
The procurement frame is what matters. Western enterprise buyers running compliance audits care about (a) data residency for non-EU jurisdictions, (b) provenance disclosure, and (c) the ability to run the model on owned hardware. Qwen 3.7 at 2M-context with open weights checks all three — except residency, which Alibaba Cloud addresses through region-specific deployments. The multilingual story is the differentiator: 201 languages is well beyond what any US-frontier lab ships natively.
The competitive read at the OSS-frontier tier is now Qwen, DeepSeek, Mistral Large 3, and Llama 5 — four families with distinct strengths. Qwen's multilingual+context combination is the most defensible. For the European enterprise buyer choosing between Mistral Large 3 (EU sovereignty story) and Qwen 3.7 (capability ceiling), the answer is increasingly "both" — split by use case.
RadarAI — China AI Models List (Updated Weekly) → · Index.dev — Top 6 Chinese AI Models Like DeepSeek (LLMs) in 2026 → · Kairntech — Top Open-Source LLMs 2026: The Best Models & Comparison →