Sulat.com
AI models
$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3 from Synthetic
DevPass (LLM Gateway) logo

Model details

Qwen3.7 Max

Qwen3.7-Max is built on a chain-of-thought reasoning architecture purpose-designed for the agent era, functioning as a versatile agent foundation rather than a pure conversational model. The model is engineered for deep agentic generalization, meaning it performs consistently whether deployed through Claude Code, OpenClaw, Qwen Code, or other scaffolding frameworks — without requiring framework-specific tuning. Its primary strengths span three axes: coding agents capable of frontend prototyping through complex multi-file engineering, office and productivity workflows powered by MCP integrations and multi-agent orchestration, and long-horizon autonomous execution that sustains coherent reasoning across hundreds or thousands of steps. The architecture's agentic focus is validated by leading benchmark scores including Terminal-Bench 2.0-Terminus at 69.7, SWE-Bench Verified at 80.4%, and SpreadSheetBench-v1 at 87.0%, placing it at the frontier of terminal coding, software engineering, and structured office automation tasks.

The model represents Alibaba's latest proprietary flagship built explicitly around agent workload demands, incorporating explicit prompt caching to reduce repeated-context overhead in long agentic pipelines. Its training lineage is positioned toward sustained autonomous execution — as demonstrated by a documented ~35-hour autonomous kernel optimization session involving over 1,000 tool calls on unseen hardware, achieving a 10.0x speedup — without degrading coherence over that extended run. Beyond pure agentic tasks, the model achieves 92.4% on GPQA Diamond and 97.1% on HMMT 2026 Feb, indicating strong scientific and mathematical reasoning foundations, while its 79.1% on IFBench reflects solid instruction-following capability. Together these results suggest Qwen3.7-Max is particularly suited for developers and enterprises seeking a single model that can credibly serve as both a coding agent and a long-horizon automation backbone, with the benchmark pedigree to back up those claims across varied agent scaffolds.

DevPass (LLM Gateway)qwen3.7-maxqwen

Quick Info

Powered by
Provider
DevPass (LLM Gateway)
Model key
qwen3.7-max
Release date
May 21, 2026
Last updated
May 21, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$1.25
Output token cost
$3.75

Limits

Output tokens
65,536 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

DevPass (LLM Gateway)

CoverageBenchmark

Fello AI's review (updated August 19, 2026) documents Qwen3.7-Max as launching on May 20, 2026 at the Alibaba Cloud Summit in Hangzhou, where Alibaba reported a 35-hour autonomous coding run that fired 1,158 tool calls and achieved a 10× geometric speedup over the Triton reference kernel. The model is positioned as Ali On benchmarks, the review confirms Qwen3.7-Max scored 56.6 on the Artificial Analysis Intelligence Index at launch (ranked 5 that week, sitting in the top 10 of 151 measured models and the highest-ranked Chinese AI model on that leaderboard to date), with API pricing of $2.50 input / $7.50 output per 1M tokens and cach

DevPass (LLM Gateway)

Coverage

Wavise OpenLLM's guide frames Qwen3.7-Max as Alibaba's flagship coding model released in May 2026 and as Alibaba's first serious proprietary coding flagship, designed for agentic workflows, long-horizon task execution, and low-level systems programming. It is closed-source, proprietary, and accessible only via API — po Specifications listed include a 1,000,000-token context window, 65,000 max output tokens, MoE design (Alibaba not fully disclosed), and native extended-thinking chain-of-thought before responding. The guide highlights 60.6% on SWE-bench Pro as the highest score among proprietary models, plus top-tier Kernel Bench perfo

Videos about Qwen3.7 Max

More models around Qwen3.7 Max