📊 Full opportunity report: The Weights Came First: What Thinking Machines’ Inkling Actually Signals on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines has released the full weights of its new model Inkling on Hugging Face under an open license, emphasizing transparency. This move signals a shift in how foundational models are shared and used, raising questions about licensing and restrictions.
Thinking Machines has publicly released the full weights of its latest foundation model, Inkling, on Hugging Face under an Apache 2.0 license. This marks a significant departure from typical industry practice, where model weights are often kept proprietary or released under restrictive licenses. The move is notable for its transparency and for challenging the conventional approach to AI model distribution, especially given Inkling’s size and capabilities.
Inkling is a 975-billion-parameter Mixture-of-Experts transformer supporting multimodal inputs, including text, images, and audio, with a shared processing space. It was trained on 45 trillion tokens across multiple modalities, with a focus on safety and performance benchmarks. The full weights were made available on Hugging Face, under an Apache 2.0 license, enabling anyone to download, modify, and deploy the model on their own infrastructure.
Unlike many recent model releases, which often provide only API access or limited weights, Inkling’s full weights are openly accessible from day one. However, reports suggest that Thinking Machines maintains a separate Model Acceptable Use Policy (AUP) that restricts certain applications, such as surveillance and fully automated decision-making affecting individuals, raising questions about the true openness of the release. The company also disclosed that the training data and pipeline are not publicly available, which is typical but limits full transparency.
The weights came first: what Inkling actually signals
Mira Murati’s lab shipped its first foundation model — and the model isn’t the story. The order of operations is: full weights, Apache 2.0, day one, before any closed API. Plus a rare concession — the lab says it’s not the strongest model available, open or closed.
- AIME 2026 97.1%
- GPQA Diamond 87.2%
- MCP Atlas (Nemotron 44.7%) 74.1%
- VoiceBench · open-weight audio frontier 91.4%
- FORTRESS adversarial · best open 78.0%
- ForecastBench · calibration 61.1
- HLE text-only (GLM-5.2 40.1%) 29.7%
- SWE-bench Pro (GLM-5.2 62.1%) 54.3%
- Terminal-Bench 2.1 (GLM-5.2 82.7%) 63.8%
- SWE-bench Verified (Fable 5 95.0%) 77.6%
- Design Arena · 2nd open, behind GLM-5.2 ~10th
A 0.2 → 0.99 effort setting trades reasoning tokens against cost & latency, so you get a curve, not a point. On Terminal-Bench 2.1 it reportedly matches Nemotron 3 Ultra at ~⅓ the tokens. Peak score is a vanity metric when you serve millions of calls; the cost curve is what ships. (Bonus: its chain of thought compressed on its own during RL — nobody rewarded it; efficiency did.)
Pitched as the Western alternative to Chinese open weights (censorship-resistance training is the differentiator). But GLM-5.2 still wins on agentic/reasoning and Kimi K2.6 often on multimodal: best American open model, second in the open field. The irony — post-training was bootstrapped on synthetic data from Kimi K2.5.
BF16 needs ≥2 TB aggregate VRAM (8× B300 / 16× H200). NVFP4 still needs ≥600 GB. Not a workstation model — a 512 GB fleet falls just short. “Open” ≠ “runnable.” Mitigations: 1-bit GGUFs (~74% acc.), hosted eval routes, and Inkling-Small (12B active) — the release local-first builders actually want.
Open weights used to be a consolation prize. Inkling is a strategic open release — Apache 2.0, natively multimodal, honestly marketed, published complete on day one, optimized for deployment rather than headlines (the model isn’t the product; the fine-tuning platform is). It doesn’t need to win every benchmark for that to matter. The frontier is learning that owning the base beats renting the API — arriving now from the inside. For the sovereignty buyer: ① a real Western hedge against being switched off · ② verify the use policy before you build · ③ check the VRAM, then benchmark vs GLM-5.2 & Kimi K2.6 on your task.
Implications of Full Weights Release for AI Transparency
This release signals a potential shift toward greater transparency and ownership in foundational AI models, allowing users to fine-tune, inspect, and deploy models independently. It challenges the industry norm of restricting model weights and raises important questions about licensing, usage restrictions, and the balance between openness and control. For organizations, this could mean more accessible tools for AI development, but also underscores the need to scrutinize licensing terms and usage policies to ensure compliance and ethical use.
Accelerate Everything with Tensor Cores: A Developer’s Guide to High-Performance AI, Efficient Training, and Scalable Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Industry Practices and the Shift Toward Open Weights
Historically, most large AI models have been released with limited access, often through APIs or with proprietary weights, partly to control misuse and partly due to competitive concerns. Recently, some companies have begun sharing weights more openly, but often with restrictions or layered policies. The case of Inkling is notable because it is a large, multimodal model with publicly available weights under an open license, yet accompanied by a separate use policy. This development comes amid broader industry debates on model transparency, ownership, and responsible AI deployment.
“Our goal is to empower developers and researchers with full access to Inkling’s weights while maintaining responsible use through our policies.”
— Thinking Machines spokesperson
high performance GPU for AI development
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Uncertainties About Usage Restrictions and Data Transparency
It remains unclear how strictly the Model Acceptable Use Policy (AUP) will be enforced and how it might impact users’ ability to modify or deploy the model freely. The training data and pipeline are not publicly disclosed, which limits full transparency and raises questions about reproducibility and bias mitigation. Additionally, the extent of restrictions on derivative works and commercial use has not been fully clarified and warrants further investigation.
AI model deployment server
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Industry Adoption and Policy Clarification
Expect further analysis from the AI community regarding the legal and ethical implications of the layered licensing approach. Companies and researchers will likely test the boundaries of the open weights and the AUP, while regulatory bodies may scrutinize the licensing terms. Additionally, more models may follow this example if the approach proves sustainable, prompting a shift in industry standards around model sharing and ownership.
machine learning model storage
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What does releasing the full weights under Apache 2.0 mean?
It means anyone can download, modify, and deploy the model freely, as long as they comply with the license terms. However, restrictions may still apply through separate policies.
How does this release differ from typical model launches?
Most large models are released via API or with limited weights; Inkling’s full weights are openly available from day one, which is uncommon at this scale.
Are there any restrictions on using Inkling?
Yes, according to reports, Thinking Machines has a separate Model Acceptable Use Policy that limits certain applications, though details are not fully confirmed.
Why is the layered approach to licensing important?
It raises questions about the true openness of the model and whether restrictions undermine the benefits of open licensing, impacting trust and usage rights.
What are the implications for AI safety and ethics?
The open release allows for greater scrutiny and research, but the restrictions and lack of data transparency could complicate efforts to evaluate bias and safety thoroughly.
Source: ThorstenMeyerAI.com
