Qwen3.7 Max

Qwen3.7-Max is positioned by the Qwen team as a proprietary foundation model tailored for the agent era, designed to serve as a versatile base for systems that need to plan, act, and persist across many steps. Rather than focusing on a single task, it is framed as a broad agent foundation that can underpin coding assistants, office productivity tools, and orchestrated multi-agent pipelines. Official documentation is available through Alibaba Cloud Model Studio, signaling first-party distribution alongside the Qwen blog announcement.

The model's practical strengths center on three areas that the launch material directly highlights. As a coding agent, it is described as comfortable with both frontend prototyping and complex multi-file engineering work. For office and productivity use, it is presented as a reliable assistant when wired into MCP-based integrations and multi-agent orchestration. Its most distinctive qualitative claim is sustained long-horizon reasoning, illustrated by a reported 35-hour fully autonomous kernel optimization run with over one thousand tool calls, along with consistent behavior across different agent scaffolds such as Claude Code, OpenClaw, and Qwen Code.

Charm Hyperqwen3.7-maxqwen

Open provider page Provider docs

Quick Info

Provider: Charm Hyper
Model key: qwen3.7-max
Release date: May 28, 2026
Last updated: Jul 22, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $2.50
Output token cost: $7.50

Limits

Output tokens: 64,000 tokens
Context window: 1,000,000 tokens

Transparent token rates

Compare Qwen3.7 Max pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Qwen3.7 Max

Charm Hyper

CoverageBenchmark

Alibaba's Qwen team has released Qwen3.7-Max, a proprietary flagship positioned at the top of the Qwen3.x family above Qwen3.6-Plus, available as a closed-source, API-accessible model through Alibaba Cloud Model Studio with a 1 million token context window and text-only input and output. According to the DataCamp write Benchmark figures cited from Qwen/Artificial Analysis put Qwen3.7-Max ahead of several frontier rivals on targeted evaluations: GPQA Diamond at 92.4 versus Claude Opus 4.6 Max's 91.3, and the Apex reasoning benchmark at 44.5 versus DeepSeek V4 Pro's 38.3. In a 35-hour autonomous kernel optimization run, the model repor

Charm Hyper

CoverageBenchmark

A Weights & Biases community report titled "Qwen3.7-Max Benchmark Scores," last updated May 20, 2026, frames Qwen3.7-Max as Alibaba's new frontier model focused on coding agents, autonomous execution, and long-horizon reasoning. The page appears within W&B's ml-news stream and lists Alibaba Qwen3 among its tracked mode As on-topic benchmark aggregation for the exact Qwen3.7-Max subject, this report is relevant corroborating material alongside the more detailed DataCamp coverage. However, because the captured excerpt shows only the title, framing, and site chrome, the specific Qwen3.7-Max scores it tabulates are not independently veri