Alibaba Releases Qwen3.8-Max: Rivaling Frontier Closed-Source Models in Coding and Multimodality

Alibaba has released its latest flagship models, Qwen 3.8 Max and a 27B version, featuring open-source weights. Initial tests by developers and researchers indicate robust performance in coding, agentic tasks, and multimodal capabilities, making it a highly cost-effective frontier open-source model.

已确认

• 编程与Agent能力: Tests by @bindureddy, @QuixiAI, and @teortaxesTex show that Qwen 3.8 excels in agentic coding tasks, performing only slightly below K3 but at half the price. @intellectronica also highlighted its excellent tool-calling abilities, making it a strong alternative to K3. An in-depth review by @葬AI noted its first-class coding and Agent capabilities, even slightly surpassing K3 in specific dimensions.
• 多模态表现: @teortaxesTex pointed out that Qwen 3.8 Max might have reached SOTA levels in image recognition and annotation, boasting strong sample efficiency that allows for perfect distillation into the 27B version. Tests by @WorldofAI also validated its frontier performance in frontend code generation and multimodal tasks.
• 开源差距缩小: Feedback shared by @dairai noted that running the model on Hermes Agent forces a reevaluation of the gap between open-source and closed-source frontier models, with open-source now closely trailing closed-source in Agent tasks.

尚未确认

• 长程任务与产品化能力: @bindureddy noted that the MoE architecture commonly used in current open-source models, while great for benchmarking, has limited active parameters and struggles with long-horizon tasks. @葬AI's review mentioned that despite excellent performance in real browser tasks, it still lags behind K3 in one-shot frontend productization, such as generating complete webpages.
• 与 K3 的绝对实力对比: @ramoscasals shared shader test results from renowned scholar Ethan Mollick, stating that while Qwen 3.8 Max is robust, it hasn't yet reached the level of Kimi K3 in his experience. @QuixiAI added that its practical performance still needs detailed comparison against models like DeepSeek.

为什么重要

The release of Qwen 3.8 Max signifies that open-source models, while maintaining high cost-effectiveness (half the price of competitors), can now rival or closely approach top-tier closed-source models in core agentic coding and multimodal capabilities. As @orange noted via feedback from their AI partner product Cola, the model's high efficiency is directly empowering downstream applications, providing developers with a more affordable, frontier alternative.

2026-08-03 ~ 2026-08-05 · 13 related posts

• Qwen 3.8 Max Released, Praised as Cost-Effective; Cola App Lauds Its Capabilities — oran_ge · 2026-08-03
• Testing Qwen3.8-Max: Open Models Are Catching Up with Closed Frontier — dair_ai · 2026-08-04
• Qwen 3.8 Max is said to lead image recognition and distill cleanly to 27B — teortaxesTex · 2026-08-04
• Qwen 3.8 Early Tests: Good at Agentic Coding, But Struggles with Long Loops — bindureddy · 2026-08-04
• [source] Qwen 3.8 Coding Test: Nearly Matches K3 at Half the Price — bindureddy · 2026-08-04
• Qwen 3.8 Max Tested: Rivals Frontier Models in Coding and Multimodal — WorldofAI · 2026-08-04
• Qwen 3 Preliminary Test: Near K3 Coding Performance at Half the Price — QuixiAI · 2026-08-04
• Developer tests Qwen3.8-max: A highly efficient and capable model — intellectronica · 2026-08-04
• Qwen 3.8 Max Tested: Solid Performance but Falls Short of Kimi K3 — ramos_casals · 2026-08-04
• [source] Qwen3.8Max Deep-Dive: Top-Tier Coding & Agent Performance at a Fraction of the Cost — 葬AI · 2026-08-05
• Building a Beautiful Invoice Generator in 10 Minutes with Alibaba's Qwen3.8-Max — CodeByPoonam · 2026-08-05
• [source] Alibaba Releases Qwen3.8-Max: 2.4T Parameters, 95B Active, Natively Multimodal — CodeByPoonam · 2026-08-05

1 near-duplicate retellings: oran_ge