Alibaba unveils new flagship "Qwen3.7-Max" with a 1M-token context window
What's happening: On May 20, 2026, the new flagship "Qwen3.7-Max" was unveiled at an Alibaba Cloud event. It is a proprietary model specialized for reasoning and agentic use that can handle a long context of up to 1 million tokens.
Key points: According to reports, it ranked near the top in independent benchmarks (a metric score of 56.6, placing 5th) and is said to have outperformed Google Gemini 3.5 Flash. Pricing is reported at $2.50 input / $7.50 output per 1M tokens β roughly half the price of top-tier Western models (all figures published or reported).
What it means: The options for using high-performance models cheaply are growing, pushing the cost competition in agent development a step further. With the rise of China-born models, choosing the right model for each use case becomes even more important.