Qwen3.7-Max

A comparative analysis (ofox.ai, June 2, 2026) confirms Qwen 3.7 Max was released on 2026-05-21 and ships with a 1M-token context window, text-only modality, and a 35-hour autonomous run ceiling supporting 1,000+ sequential tool calls. Both Qwen 3.7 Max and the sibling Qwen 3.7 Plus (released June 1, 2026) share these On coding-agent benchmarks, Qwen 3.7 Max holds the SWE-Bench Pro edge at 60.6%, but the article argues that for most coding, document, and agent workloads Qwen 3.7 Plus offers a more cost-effective choice given only a roughly 2-point benchmark differential. Max retains a modest latency advantage (7–15% faster cold path

The explainx.ai blog post dated May 22, 2026 frames Qwen 3.7-Max as Alibaba's "versatile agent foundation," unveiled on May 20, 2026 and served via Alibaba Cloud Model Studio, with a stated design goal of sustaining coherent reasoning across extremely long horizons rather than maximizing abstract reasoning test scores. The supplied excerpt's coding-agents benchmark table shows Qwen 3.7-Max scoring 69.7 on Terminal-Bench 2.0 (vs Opus 4.6 Max 65.4 and DeepSeek V4 Pro 67.9), 80.4 on SWE-Bench Verified, 60.6 on SWE-Bench Pro, and 53.5 on SciCode — corroborating the Artificial Analysis and SWE numbers from other on-subject candidates. Per

DataCamp's profile describes Qwen3.7-Max as Alibaba's proprietary flagship positioned above Qwen3.6-Plus, with a 1 million token context window, text-only input/output, and availability through Alibaba Cloud Model Studio. It cites an Artificial Analysis Intelligence Index score of 56.6 and benchmark results including 9 The same report flags notable verbosity — approximately 97 million tokens generated during Artificial Analysis evaluation versus a 24 million median — which has cost implications for long agentic sessions. The piece positions Qwen3.7-Max as a closed-source, API-accessible model designed as a "versatile agent foundation

The llm-stats.com scorecard for Qwen3.7-Max, last refreshed around September 10, 2026, places the model at an overall composite rank of 26 (LLM Stats Score 45.6) at a blended price of roughly $1.37 per million tokens, and shows it holding strong in vertical domains — top 2% in Finance (1 of 227), Legal (1 of 210), and The supplied excerpt also reports conversation-depth stability: a Turn 1 score of 15.4 (12 of 114) that holds at 15.5 for Turns 31+ (2 of 80), with a quality tracker marked Stable (+0.72σ across 337 votes, +1.91σ on Chat), suggesting Qwen3.7-Max maintains coherent output across long sessions. Per-dataset entries in the

The Weights & Biases ml-news report aggregates Qwen3.7-Max benchmark scores and frames it as a frontier AI model focused on coding agents, autonomous execution, and long-horizon reasoning from Alibaba's Qwen team. The excerpt identifies the report's author (Brett Young) and community-publication context within W&B's ml The supplied excerpt is truncated before the actual benchmark score tables, so no specific Qwen3.7-Max numbers can be verified from this page alone. No NovitaAI-specific provider details (pricing, latency, throughput, uptime) for Qwen3.7-Max appear in the supplied text. The evidence supports only that a W&B community r