Alibaba (Qwen Team)
Alibaba's flagship agentic model — 1M context, 35-hour autonomous tasks
Qwen 3.7 Max is Alibaba's most capable model as of May 2026, built for long-horizon agentic execution — autonomously running for up to 35 hours across thousands of tool calls and code iterations without human intervention. With a 1M-token context window and 65,536 max output tokens, it ranks #1 on BenchLM's leaderboard (score 92, ahead of DeepSeek V4 Pro at 88). Available via API on Alibaba Cloud, OpenRouter, and Together AI; open weights expected around June 2026. **Step-by-step (API):** 1) Sign up at chat.qwen.ai for free preview. 2) For API, access via OpenRouter (`qwen/qwen3.7-max`) or Alibaba Cloud DashScope. 3) Use `enable_thinking: true` for full reasoning mode or `false` for fast responses. 4) Best for: autonomous coding agents, long research pipelines, multi-step business workflow automation.
Autonomously executes complex tasks for up to 35 hours with 1,000+ tool calls.
API: $2.50/M input · $7.50/M output · Free preview at chat.qwen.ai