Latest open artifacts (#24): Motif-3, GLM

Avid Artifacts readers know that we have been covering not only models but also their licenses for quite some time. There was a period when custom licenses were all the rage, for example the custom Qwen2.5 72B-Instruct license or the Llama licenses. DeepSeek had a custom license for DeepSeek V3 before R1 changed it to MIT, which has resulted in many (Chinese) model makers adopting MIT or Apache 2.0 licenses in 2025.
In 2026, open models are more competitive than ever, which has led to two interesting developments: Western model makers adopt open licenses, with both Google and Meta switching to Apache 2.0. Chinese model makers at the frontier, however, are becoming more restrictive: Kimi K3 comes with a license which requires commercial agreements for those who run inference or fine-tuning services, and MiniMax M3 requires agreements above a revenue threshold and has prohibited use cases.
The newest addition is Zhipu’s GLM-5.3, which switched from MIT (GLM-5.2 and earlier) to a custom license with the following clause for inference and fine-tuning providers:
If the Licensee or any of its affiliates operates a Model as a Service business, and the aggregate revenue of the Licensee and its affiliates exceeds 10 billion US dollars (or the equivalent in other currencies) in total over any consecutive 12 months, the Licensee must pass Z.AI’s security review before using the Software or its derivative works for any commercial purpose. The scope and method of the security review shall be reasonably determined by Z.AI.
While the 10 billion US dollar threshold is very high compared to other licenses of this kind, “affiliates” is not defined in the license, which adds uncertainty and creates barriers to adoption. Furthermore, the license is provided in both English and Chinese, with the Chinese text using “关联方” for affiliated parties, which does have a definition in Chinese law.
We are by no means legal experts and there are obvious reasons why those licenses are created. However, we want to highlight the issues that come with creating such licenses, especially in a world with a lot of valid open and closed alternatives.
Our Picks
Motif-3 by Motif-Technologies: Motif is one of the few hidden gems out there, showcasing innovation in their model training with very limited resources compared to others. Motif-3 comes with an MIT license and impressive scores for its size. Given the trajectory of model releases from Motif 2.6B, which we covered in 2025 and Motif-2-12.7B, the improvements are impressive.
dots3-note-prev by dots-studio: RedNote/Xiaohongshu, the Chinese Instagram, is also getting more serious about model training, although they aren’t exactly a newcomer, having released models as early as 2025. dots3 was also able to win the IMO 2026 with a perfect score using an internal harness. We expect more from them in the near future.
Qwen3.8-Flash-Next by Qwen: A preview of the next version of Qwen models in terms of architecture: 125B-A6B with 51B n-gram embeddings. It uses GDN and Qwen Sparse Attention. Similar to Qwen3-Next-80B-A3B-Instruct, we expect similar architectures to become more popular and the ecosystem to fix integrations by the time Qwen4 drops.
GLM-5.3-Flash by zai-org: This release perfected the version of the Chinese model playbook we’ve written about in 2025: The model got released as a free-to-use “stealth model” under the name “Ox-Alpha” on OpenRouter and OpenCode, which got people excited to try it out in the first place. They then speculated about its creator and size, alleging it is a >1T model from Cursor/xAI, Gemini or a new pre-train from open source labs. Because the model is relatively performant, people kept speculating for days about its creator, thus building up hype. It also dampens the accusations of benchmaxxing which accompany every (open) model release.
Hy4-preview by tencent: Tencent is becoming a serious player in the open model space, increasing the size of their flagship model while spinning the post-training flywheel. The result, Hy4-preview, is a competent model which currently has an issue with overthinking. However, if the trajectory from Hy3-preview to Hy3 is any indication, the final model might be a legit shot at the front ranks of open models.
View more details on all the models in this issue at our Artifacts Hub.
Models
General Purpose
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 by nvidia: An update to Nemotron, which comes with performance — but especially speed improvements — across the board.
Ling-3.0-flash by inclusionAI: Ant Ling is a frequent guest at the Artifacts Log; they are now on their third iteration of models, adopting a hybrid design (KDA + Gated MLA), similar to others. They also release a small 7.9B-A1.3B version.
Qwen3.8-2.4T-A95B by Qwen: In a rather surprising turn of events, Alibaba started to openly release their biggest versions of Qwen as well. However, it comes with a custom license and its performance is behind other models of its size.

Facts Only

* Custom licenses were popular for models such as Qwen2.5 72B-Instruct and Llama.
* DeepSeek had a custom license for DeepSeek V3 before switching to MIT, which influenced many Chinese model makers to adopt MIT or Apache 2.0 in 2025.
* Western model makers, including Google and Meta, are adopting Apache 2.0 licenses.
* Chinese model makers are becoming more restrictive with models like Kimi K3, requiring commercial agreements for inference or fine-tuning services.
* MiniMax M3 requires agreements above a revenue threshold and has prohibited use cases.
* Zhipu’s GLM-5.3 switched from MIT to a custom license with a security review clause for Model as a Service providers earning over $10 billion annually from affiliates.
* The GLM-5.3 license uses the term “affiliates,” which lacks a definition, creating uncertainty.
* The license for GLM-5.3 is provided in both English and Chinese, with the Chinese text using “关联方” (affiliated parties).
* Motif-3 has an MIT license.
* Qwen released Qwen3.8-Flash-Next, an architecture preview of 125B-A6B.
* Alibaba openly released its largest Qwen versions with a custom license.

Executive Summary

The landscape of open and closed model licensing is evolving, driven by competitive dynamics between different geographic regions. Historically, custom licenses were popular for models like Qwen2.5 72B-Instruct and Llama. A shift is occurring as Western developers, including Google and Meta, adopt more permissive licenses like Apache 2.0. Conversely, Chinese model makers are introducing more restrictive licensing terms. For instance, Kimi K3 requires commercial agreements for inference or fine-tuning services, and MiniMax M3 imposes revenue thresholds and prohibited use cases. A new standard emerges with Zhipu’s GLM-5.3, which mandates a security review for inference and fine-tuning providers operating Model as a Service businesses exceeding a $10 billion annual revenue threshold from affiliates. This introduces licensing uncertainty due to undefined terms like "affiliates," though the license is provided in both English and Chinese. Meanwhile, new model releases show diverse practices, with Motif-3 using MIT, Qwen releasing models under custom licenses, and Tencent developing larger models.

Full Take

The narrative reveals a tension between the global trend toward open licensing and specific regional strategies regarding commercial control over proprietary assets in the rapidly evolving AI ecosystem. The shift from permissive licenses to restrictive custom agreements, exemplified by Zhipu’s high-threshold requirement, suggests an emerging paradigm where large-scale commercial deployment triggers regulatory or security oversight. This dynamic challenges the established notion of "openness," implying that license terms are not merely legal formalities but instruments for controlling access and establishing new economic hierarchies between model creators and service providers. The ambiguity introduced by undefined terms like “affiliates” demonstrates how legal language can be weaponized to create friction and uncertainty, thereby creating artificial barriers to adoption rather than facilitating collaboration. The performance data presented—from Motif’s efficiency to the architectural updates in Qwen—suggests that innovation is proceeding rapidly, but the governance structure around this innovation is being selectively tailored based on geopolitical context. The implications point toward a fragmentation where open source frameworks coexist uneasily with bespoke contractual regimes, forcing users to navigate not just technical capability but complex legal risk profiles when engaging with frontier models.

Sentinel — Human

Confidence

This text reads like an expert editorial or deep-dive report combining technical model analysis with regulatory/licensing speculation, strongly suggesting human authorship rooted in specific industry knowledge.

Signals Detected

Sentence length variance is erratic; strong use of parenthetical comments and conversational phrasing.

Features idiosyncratic framing ('Avid Artifacts readers know that...') and transitions between technical details and speculative commentary.

The text flows as a curated editorial piece, referencing specific internal events (e.g., 2025/2026 developments) typical of specialized long-form journalism.

The structure suggests synthesis based on pre-existing, highly specific internal knowledge or established reporting structures rather than pure LLM generation.

Human Indicators

Use of highly specific, time-bound references (e.g., Qwen2.5 72B-Instruct license, Motif-3 performance trajectory) indicative of deep domain expertise.

The blending of legal analysis (license clauses) with product hype creates a nuanced, opinionated tone.