Logo
Qwen · Qwen3.7 Max · May 2026

Qwen3.7 Max — flagship of the Qwen series

  • Million-token context
  • 65K max output
  • Hybrid thinking on/off
  • Tools & structured output
Qwen3.7 Max

What is Qwen3.7 Max?

Qwen3.7 Max is Alibaba Cloud Tongyi Qianwen’s Qwen3.7 text flagship from around May 2026, for Agents, coding, office work, and hard reasoning. Versus Qwen3.6 Max and earlier Max tiers, context rises to a million tokens, with hybrid thinking and tool use emphasized. For vision input, use Qwen3.7 Plus. Pick Qwen3.7 Max in iMini Agent to use it.

Vendor
Qwen
Released
May 2026
Context
1M tokens
Max output
65K tokens
Deep thinking
Hybrid thinking · on/off
Best as
Qwen peak-reasoning workhorse

What’s new vs Qwen3.6 Max

Context, hybrid thinking, and Agent tooling—the changes to check before you make Max your workhorse.

Context up to a million tokens

Versus the common 256K tier on 3.6, long specs and large repos can stay in one thread.

Hybrid thinking on/off

Turn thinking on when needed; turn it off on simple jobs for speed; keep depth on hard ones.

Fuller tools and structured output

Function calling, built-in tools, and structured output for business systems and Agent flows.

Agent-facing headline benches

Official focus on Terminal-Bench, SWE, and similar Agent suites—fit as a coding and automation workhorse.

Official benchmarks

Official charts from the Qwen team’s “Qwen3.7: The Agent Frontier.”

Qwen3.7-Max aggregate scoreboard: coding Agents, general Agents, reasoning, and multilingual
Aggregate scoreboard. Coding Agents (SWE-Pro, Terminal Bench, and more), general Agents (MCP-Mark, MCP-Atlas, SkillsBench), reasoning (GPQA, HLE, HMMT), plus instruction/multilingual—shown with Opus, DeepSeek V4 Pro, GLM, and peers. Use it to place Max in Agent settings.
Qwen3.7-Max Agent environment-scaling trajectory and rank changes
Agent environment-scaling trajectory. As training environment quality and diversity rise, average rank moves toward Claude Opus 4.6 Max—showing what environment scaling contributes, not a one-shot leaderboard spike.
Qwen3.7-Max cross-harness generalization benches
Cross-harness generalization. The same model on Claude Code, OpenClaw, Qwen Code, and other scaffolds—whether it only fits one framework or holds across scaffolds.
Qwen3.7-Max reward-hacking self-monitoring in long-horizon training
Reward-hacking self-monitoring in long-horizon RL. The model finds and flags score-gaming in SWE RL—about training stability, not everyday chat scores.

Three common workflows

Qwen3.7 Max fits Qwen jobs that need peak text reasoning.

Hard coding and Agents

Multi-step coding and tool orchestration—as Qwen’s text flagship.

Office and decision materials

Long-document synthesis, table reasoning, and one-page memos.

Hard-constraint reasoning

Derivations with clear rules and numbers—keep steps checkable.

Vs Qwen3.7 Plus and DeepSeek V4 Pro

All three sit at peak or flagship. Gaps are vision, product stack, and open-weight path.

DimensionQwen3.7 MaxQwen3.7 PlusDeepSeek V4 Pro
Context1M tokens1M tokens1M tokens
ReasoningHybrid thinking · peakHybrid thinking · balancedThink high / Max
Coding / AgentsQwen text flagshipBalanced pick with visionOpen hard-coding flagship
Long tool runsTools & structured outputBuilt-in tools + imagesLong-horizon automation focus
Speed posturePeak quality firstPerformance/cost more balancedCompletion quality first
Prefer whenQwen peak text reasoningEveryday Qwen multimodal workhorseDeepSeek code & hard reasoning

Using Qwen3.7 Max on iMini

Free quota, model identity, key specs, and how it differs from Plus.

More models

More flagship models by vendor

GPT-5.6 SolOpenAI’s depth flagship—when the team is on the GPT stack and needs maximum depth.Claude Opus 5Anthropic’s Opus flagship—for long-horizon coding and multi-step Agents on the Claude stack.Gemini 3.1 ProGoogle’s Gemini Pro tier—for multimodal, long-context work on the Google stack.Grok 4.5xAI’s Grok flagship—when you need Grok reasoning or the xAI product stack.DeepSeek V4 ProDeepSeek’s coding and reasoning flagship—a strong open-weight pick for code and hard reasoning.Kimi K3Moonshot’s Kimi flagship—for CN/EN long context and China-market delivery.GLM 5.2Zhipu’s GLM flagship—for bilingual Agents and Zhipu-stack delivery.MiniMax M3MiniMax’s flagship—when delivery is tied to the MiniMax stack.Llama 4 MaverickMeta’s Llama 4 flagship option—when you need the Llama weight stack.Seed 2.0 LiteByteDance Seed’s lighter tier—for high-turnaround, shorter sessions.
All models

Use Qwen3.7 Max free on iMini Agent

Million-token context and hybrid thinking—for peak reasoning and Agents.

Open Qwen3.7 Max

No Qwen API key required.