📊 Full opportunity report: Decoding The Inkling: What It Means For AI’s Next Chapter on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines Lab has released Inkling, a 975-billion-parameter open-weight AI model, emphasizing transparency and open access. The model’s specifications and licensing are confirmed, but some claims about policies and data remain uncertain. This development signals a shift toward more open, yet regulated, AI models.
Thinking Machines Lab has officially released Inkling, a 975-billion-parameter multimodal AI model, under an open-source license, marking a significant step in transparency and accessibility for large language models. This move, involving the immediate availability of full weights on Hugging Face, underscores a shift in how AI models are shared and owned, with implications for the industry’s approach to open-source development and control.
Inkling is a Mixture-of-Experts transformer with 975 billion total parameters and 41 billion active, supporting a 1-million-token context window. It was trained on 45 trillion tokens across text, images, audio, and video, with a native multimodal input design that processes text, images, and audio jointly without relying on vision adapters. The model is available under the Apache 2.0 license, allowing download, modification, and commercial use, with full weights accessible on Hugging Face.
However, the release is accompanied by caveats. The weights are not open source in the strictest sense; the training data and pipeline are not publicly disclosed. Additionally, reports suggest Thinking Machines maintains a separate Model Acceptable Use Policy (AUP) that restricts surveillance, deception, and automated decision-making, raising questions about the scope of open access and restrictions beyond licensing.
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-Weight Release for AI Ownership
This release signifies a move toward more transparent and owner-controlled AI models, enabling organizations to fine-tune, inspect, and deploy models independently. It challenges the industry norm of proprietary models by offering full weights under an open license, potentially accelerating innovation and democratization of AI technology. Still, the existence of additional use restrictions through a separate policy complicates the narrative of true openness, highlighting ongoing tensions between openness and control in AI development.
Yahboom Raspberry Pi 5 ROS2 Robot Car 360°Movement, AI Vision & Tracking, Integrated Multimodal Large AI Model OpenRouter, AI Voice Interaction (Superior Without RPi5)
- Powerful Raspberry Pi 5 Control: Enhanced processing, multimedia, and AI performance
- Large AI Model Integration: Advanced human-computer interaction and environmental perception
- Multiple Control Options: App, PC, remote, and handle control with FPV
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Industry Shift Toward Transparent AI Model Sharing
Until now, most large AI models have been released with closed weights or limited access, often accompanied by licensing restrictions. The recent release of Inkling by Thinking Machines Lab marks a notable departure, emphasizing transparency and open access. The model’s release follows a broader industry trend toward open-source AI, but with nuanced restrictions that reflect ongoing debates about responsible AI use and control.
Previous major models, such as GPT-3 or PaLM, have been kept proprietary or partially open, often with licensing restrictions. Inkling’s immediate availability of full weights under Apache 2.0, combined with candid details about training and performance, signals a potential shift in how industry players approach model sharing and ownership.
“The release of Inkling under Apache 2.0, with full weights available, is a significant step toward democratizing AI development, but the additional use restrictions complicate the narrative of true openness.”
— Thorsten Meyer, AI researcher
Unclear Aspects of Inkling’s Usage Restrictions
It is still unclear how enforceable the separate Model Acceptable Use Policy (AUP) is, and whether it significantly limits the open-license benefits. The specifics of data privacy, surveillance restrictions, and the scope of modifications are not fully verified. Additionally, the implications of the model’s non-disclosed training data remain uncertain, raising questions about reproducibility and transparency.
Next Steps for Industry Adoption and Evaluation
Further independent testing and benchmarking of Inkling will clarify its real-world performance and safety features. Industry stakeholders are likely to scrutinize the AUP and licensing terms closely before adopting or building upon the model. Additionally, other organizations may follow suit, either by releasing their own open-weight models or by challenging the balance between openness and restrictions.
Key Questions
What makes Inkling different from other AI models?
Inkling is notable for its large size (975 billion parameters), native multimodal input support, immediate availability of full weights under an open license, and transparent training details, marking a shift toward more open model sharing.
Are the weights truly open source?
The weights are available under Apache 2.0 license, allowing modification and commercial use, but the training data and pipeline are not disclosed. Additionally, a separate use policy may impose restrictions, complicating the open-source classification.
What are the potential risks of this open-weight release?
Potential risks include misuse due to insufficient restrictions, challenges in verifying training data for bias or privacy issues, and uncertainties about the enforceability of additional use restrictions.
How does this affect the future of AI development?
This move could accelerate democratization and innovation in AI by making powerful models more accessible, but it also raises questions about regulation, responsible use, and the balance between openness and control.
Source: ThorstenMeyerAI.com
