📊 Full opportunity report: Exploring The Meaning Of Thinking Machines’ Inkling In AI Advancements on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines unveiled its Inkling AI model, available openly on Hugging Face under Apache 2.0 license. The move emphasizes transparency but raises questions about restrictions and data sources. The development highlights ongoing debates around open AI models and ownership costs.
Thinking Machines has officially released its Inkling model with open weights on Hugging Face, marking a notable shift in AI model accessibility. Unlike typical releases that limit access or impose restrictions, Inkling’s weights are available under the Apache 2.0 license, allowing users to download, modify, and deploy independently. This move is significant because it directly addresses ongoing debates about AI ownership, openness, and the costs associated with controlling proprietary models.
The Inkling model is a 975-billion-parameter mixture-of-experts transformer supporting multimodal inputs — text, images, and audio — with a 1-million-token context window. It was trained on 45 trillion tokens across various media types, using a hybrid optimizer and NVIDIA hardware. The model was not released with a closed API but instead shared openly on Hugging Face, with full weights accessible for download under the Apache 2.0 license. This allows organizations to fine-tune, inspect, and deploy the model on their own infrastructure, even if the company’s relationship with Thinking Machines changes.
While the weights are openly available, there are important caveats: the training data and pipeline are not public, and reports suggest that Thinking Machines maintains a separate Model Acceptable Use Policy (AUP) that restricts surveillance, deception, and automated decision-making affecting individuals’ rights. This layered policy introduces potential restrictions beyond the Apache license, which could influence how the model is used in practice.
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 Open Release and Licensing Restrictions
This release underscores a shift toward greater transparency in AI development, enabling organizations to own and control their models without reliance on proprietary APIs. It also highlights the ongoing tension between open licensing and usage restrictions, which could influence how models are adopted in sensitive domains like public safety or surveillance. The move may set a precedent for other AI labs to follow, emphasizing openness but with layered policies that could limit certain applications.
AI Systems Performance Engineering: Optimizing Model Training and Inference Workloads with GPUs, CUDA, and PyTorch
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on Open-Source AI Model Releases
Over recent years, there has been a growing movement toward open-sourcing large AI models, driven by calls for transparency, reproducibility, and democratization of AI technology. Companies like Meta, OpenAI, and others have released models with varying degrees of openness, often accompanied by licensing restrictions or usage policies. Thinking Machines’ approach with Inkling marks a departure from typical proprietary releases, emphasizing open weights but also layered restrictions. The release comes amid broader debates about the ownership, control, and ethical use of powerful AI models, especially as their capabilities continue to advance rapidly.
“We believe in empowering developers and organizations to own and adapt our models fully, which is why we released Inkling’s weights openly.”
— Thinking Machines spokesperson
multimodal AI development kits
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Usage Restrictions and Data Sources
It remains unclear how the separate Model Acceptable Use Policy will be enforced in practice and whether it will significantly limit the model’s deployment in sensitive areas. Additionally, the specifics of the training data and pipeline are not publicly available, raising questions about transparency and reproducibility. The impact of layered restrictions alongside an open license is still developing and could influence how the model is adopted and regulated.
large language model fine-tuning tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Adoption, Testing, and Policy Clarification
Organizations and developers will likely begin testing Inkling’s capabilities and assessing the practical implications of its licensing and restrictions. Further disclosures about the AUP and training data are expected, which will clarify the model’s usability in various domains. Monitoring how regulators and industry groups respond to this layered openness model will also be key in shaping future AI licensing practices.
AI model deployment infrastructure
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What makes Inkling different from other AI models?
Inkling is released with open weights under Apache 2.0, allowing full ownership and modification, but it is accompanied by a layered acceptable use policy that may restrict certain applications, making it distinct from fully open-source models.
Can I freely use Inkling for commercial purposes?
While the weights are openly available for commercial use under Apache 2.0, the accompanying AUP may impose restrictions, so users should review the policy carefully before deploying in sensitive or regulated domains.
Why is the layered restriction approach significant?
It introduces a nuanced model of openness—while the weights are accessible, the added restrictions could limit certain applications, especially in areas like surveillance or automated decision-making, raising questions about transparency and control.
Will the training data and pipeline be released?
No, the training data and full pipeline are not publicly disclosed, which limits full transparency and reproducibility of the model’s development process.
What is the significance of the open weights being on Hugging Face?
Hosting on Hugging Face makes the model easily accessible to a broad community of developers and researchers, fostering experimentation, fine-tuning, and independent deployment, which can accelerate AI innovation and ownership.
Source: ThorstenMeyerAI.com
