FLAGSHIP PREVIEW -- QWEN3.8-MAX (2026-07-19 at WAIC Shanghai, press-reported): Alibaba previewed **Qwen3.8-Max**, a **2.4-trillion-parameter sparse-MoE multimodal model** (text/image/video/docs), days after Moonshot's Kimi K3 open-weights launch. Alibaba positions it as 'second only to [Claude] Fable 5.' Access at preview is CLOSED: via Alibaba's Token Plan subscription plus the Qoder / QoderWork agentic platforms, at roughly 10% of standard pricing; **open weights 'promised soon' with no date or license, and no model card, benchmark table, or third-party eval yet** -- treat the ranking claim as unverified vendor positioning. SEPARATELY (2026-07-21) the Qwen team released **Qwen-Image-3.0**, a 3rd-gen image model (up to ~4,500-token prompts, legible text down to ~10px, 12 languages, UI/web + live-data rendering) -- but a transparency regression: **Qwen Chat access only, no open weights, no benchmarks, no technical report** (a departure from Qwen's prior Apache-2.0 same-day-weights image releases)
2026-07-19•MarkTechPost (2026-07-19, Qwen3.8-Max preview), Unite.ai (2026-07-21, Qwen-Image-3.0)
USER-CREATED AI AGENTS SHUT DOWN -- TOOK EFFECT 2026-07-15 (confirmed): Alibaba's Qwen app (and ByteDance's Doubao, simultaneously) **shut down user-created/custom AI agent features on July 15, 2026** as China's rules on 'humanlike AI interaction services' (the anthropomorphic-AI/companion rules) took effect that day; Qwen had already disabled humanlike interactive agents ~7/10 ahead of the broader cutoff. IMPORTANT DATA DETAIL: the platforms diverge -- **Doubao** offers read-only access to old agent data until **October 15, 2026** (unrecoverable after), but **Qwen offers no migration path at all**: agent configurations and conversation histories are permanently deleted. This is a consumer-app change -- the Qwen models/API are unaffected. Separately (SCMP, early July, press-reported): Alibaba cut Qwen3.7-Max by ~80% and Qwen3.7-Plus by ~60% for international users on the Qoder platform during off-peak hours (10pm-8am Beijing = US working hours) -- an explicit play for US developer demand amid a broader Chinese token price war
2026-07-15•SCMP (scmp.com/tech/big-tech/article/3359482), TechTimes (2026-07-15, effective-date confirmation), Bloomberg (2026-07-06), TechNode
FLAGSHIP GA + VENDOR NUMBERS CONFIRMED (verified 2026-06-09): **Qwen 3.7 Max** shipped 2026-05-20 ('Qwen3.7: The Agent Frontier' on the official Qwen blog) -- the benchmark figures that circulated in aggregators are now vendor-published: **SWE-Bench Pro 60.6%, Terminal-Bench 2.0-Terminus 69.7%, GPQA 92.4%**, plus SWE-Multilingual 78.3%, SciCode 53.5%, QwenSVG 1608. PROPRIETARY / API-only on DashScope/Model Studio. Official pricing: **$2.50/M input, $7.50/M output list -- with a 50% launch promo ($1.25/$3.75) running until 2026-06-22**. Separately, **Qwen3.7-Plus** launched ~2026-06-02 on Bailian/Model Studio: multimodal (image + video input, text output) with GUI-agent, deep reasoning, and tool invocation at $0.40/$1.60 per 1M -- press-reported agent scores (ScreenSpot Pro 79.0, AndroidWorld 81.0) await vendor tables. Both models confirmed on Alibaba's official Model Studio model list
2026-06-09•Qwen blog (qwen.ai/blog?id=qwen3.7), Alibaba Cloud pricing campaign (alibabacloud.com/campaign/qwen-discount), Alibaba Model Studio model list (help.aliyun.com/zh/model-studio/models), VentureBeat (Qwen3.7-Plus)
INTERPRETABILITY RELEASE (2026-05-01): Alibaba released **Qwen-Scope** -- an open-source suite of **sparse autoencoder (SAE) weights** for the Qwen3 and Qwen3.5 model families, plus inference + visualization tooling. SAEs decompose internal model activations into interpretable feature directions (Anthropic / Google DeepMind have been the highest-profile labs publishing this style of interpretability work; Alibaba is the first major Chinese lab to ship SAE weights openly). Practical implication: researchers and red-teamers can now probe Qwen3.5 internals (feature attribution, concept steering, refusal-circuit analysis) without training their own SAEs -- a multi-month / multi-GPU effort. Adds to Qwen's open-weights story even as the Qwen 3.6 flagship line goes closed-weights. Materially raises the floor for safety / alignment research on Chinese open-weight families.
2026-05-01•Qwen blog (qwen.ai/blog?id=qwen-scope)
Qwen3.6-27B dropped 2026-04-22 (~48h after Max-Preview) as the new open-weights coding champion. DENSE 27B, NOT MoE -- all params active, simpler to deploy. Apache 2.0 open weights with no commercial restrictions. Multimodal: text + image + video. 262K native context (extensible to ~1M). Verified benchmarks: SWE-bench Verified 77.2%, SWE-bench Pro 53.5%, Terminal-Bench 2.0 59.3% (matches Claude Opus 4.5 exactly), MMLU-Pro 86.2%, GPQA Diamond 87.8%, AIME 2026 94.1%, MMMU-Pro 75.8%. Notably BEATS the Qwen3.5-397B-A17B MoE flagship on coding from a single consumer GPU -- this displaces the 35B-A3B as the open-weights centerpiece for most users. Quants from unsloth + mlx-community shipped same week
2026-04-22•qwen.ai/blog?id=qwen3.6-27b, HuggingFace Qwen/Qwen3.6-27B, implicator.ai analysis
Qwen 3.6-Max-Preview launched 2026-04-20 as Alibaba's first flagship shipped WITHOUT open weights -- #1 across SWE-bench Pro, Terminal-Bench 2.0, SciCode, QwenClawBench, and QwenWebBench. 256K context, text-only (no vision at launch), APIs compatible with OpenAI + Anthropic SDKs for easy swap-in. Signals Alibaba's pivot toward closed flagships while keeping mid-size models open. If you picked Qwen for the 'best open weights' story, the best model is no longer open -- but the 3.6-27B Apache 2.0 dense released 48 hours later mostly closes that complaint for non-frontier coding workloads
2026-04•Decrypt, CNTechPost
Issues here are sourced from our editorial sweeps, not real-time telemetry. Newer issues may exist.