Model

Qwen3.7-Max NO.32

Alibaba's proprietary agentic reasoning flagship with a 1M context.

Alibaba Qwen · Qwen · Balanced · Closed weights

Specification

Fig.01Qwen — Alibaba Qwen

CHINA

Context window
1Mtokens
Max output
66Ktokens
Input price
$2.50/ 1M
Output price
$7.50/ 1M
Cached input
$0.25/ 1M
Throughput
Class
Balanced
Modalities
Text
Weights
Closed
Knowledge cutoff
Dec 2025
Released
May 2026

Overview

Qwen3.7-Max is Alibaba's current proprietary flagship (May 2026, the successor to Qwen3-Max), an agent-oriented reasoning model with a 1M-token context window and an extended-thinking mode aimed at long-horizon tasks. It posts leaderboard-topping numbers and is exceptional value — list pricing is $2.50/$7.50 per 1M tokens, and Alibaba frequently runs a 50%-off promo — with genuinely strong multilingual and math ability. Those benchmarks overstate day-to-day dependability, though — on long autonomous coding it is clearly a step below the Claude/GPT frontier, and it is text-only.

Strengths

  • Outstanding multilingual coverage
  • Strong math and reasoning benchmarks
  • 1M-token context for long tasks
  • Aggressive pricing for a flagship

Trade-offs

  • Real-world agentic coding below Claude/GPT
  • Benchmark scores overstate reliability
  • No native vision in the Max tier
  • Can loop or over-plan on long agent runs

Fit

Best for

  • Multilingual reasoning and generation
  • Math and analytical tasks
  • Cost-sensitive agentic workflows
  • Long-context document processing

Not ideal for

  • Mission-critical autonomous coding
  • Vision-dependent tasks

Capabilities

Ten axes, normalized 0–100 and scored the same way across the whole catalog — so a 78 here means what a 78 means anywhere else on the bench.

Fig.02Capability radar

0 — 100

Reasoning
85
Coding
76
Math
90
Writing
82
Knowledge
86
Speed
62
Agentic
74
Vision
5
Multilingual
92
Long Context
84

Benchmarks

Public results, with independent third-party runs marked. Bars normalize percentages against 100 and Elo ratings against a 1500 ceiling.

Independent index

Artificial Analysis Intelligence Index

Composite of ~9–10 independent evals · Artificial Analysis

24/ 100

BenchmarkResult
Artificial Analysis Intelligence IndexIndependentGeneral · Artificial Analysis
24%
SWE-bench VerifiedIndependentCoding · SWE-bench
69.6%
τ-benchIndependentAgentic · τ-bench
74.8%
LMArena EloIndependentGeneral · LMArena
1428Elo

Alternatives

Models in roughly the same class — the ones worth weighing against this record.

Index

All models/EDUCATION