📊 Full opportunity report: The First Clues From Thinking Machines That Could Shape AI’s Future on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines has released the full weights of its new multimodal AI model Inkling under an open license. This move emphasizes transparency and ownership, but raises questions about restrictions and data use. The development signals a shift toward more openly available foundational models in AI.
Thinking Machines has released the full weights of its new multimodal AI model Inkling under an open-source license, making it one of the first foundation models of its kind to do so. This move challenges industry norms by prioritizing transparency and ownership over closed APIs or commercial exclusivity, and it could influence future AI development and deployment strategies.
The Inkling model is a 975-billion-parameter mixture-of-experts transformer supporting a 1-million-token context window. It was pretrained on 45 trillion tokens across text, images, audio, and video, with native multimodal input capabilities for text, images, and audio, all processed jointly without an encoder. The full weights were published on Hugging Face under Apache 2.0 license, allowing download, modification, and deployment by anyone, even for commercial use.
Unlike typical industry practice, the release was not accompanied by a closed API or restricted access. The lab explicitly acknowledged that Inkling is not the strongest model available, emphasizing honesty over hype. The release also included a smaller variant, Inkling-Small, with 276 billion total parameters, which performs competitively on benchmarks.
However, there are important caveats: the weights are not open source in the traditional sense, as the training data and pipeline are not published. Additionally, reports suggest that Thinking Machines maintains a separate Model Acceptable Use Policy that restricts surveillance, deception, and certain decision-making applications, raising questions about the scope of the open license.
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 Weights Model Release
This release marks a significant shift toward greater transparency and ownership in foundational AI models. By providing full weights under an open license, Thinking Machines enables organizations to fine-tune, inspect, and deploy the model independently, potentially accelerating innovation and reducing reliance on proprietary APIs. It also raises important questions about ethical use, licensing restrictions, and data provenance, especially given the reported separate use policies that may limit certain applications. Overall, this move could influence industry standards for model openness and ownership.
AI Engineering: Building Applications with Foundation Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on AI Model Releases and Industry Norms
Traditionally, major AI companies have favored closed models with restricted access, often providing only API-based usage to control deployment and monetize their technology. Open-source releases are rare and usually involve limited weights or smaller models, partly due to concerns over misuse, data privacy, and intellectual property. Recent developments, including Meta’s Llama and OpenAI’s GPT models, have shown some openness, but full weights are seldom released. The release of Inkling’s full weights under Apache 2.0 is a notable departure, emphasizing transparency and user ownership.
In the past year, there has been increasing debate about the balance between openness and safety, with some advocates pushing for more accessible models to democratize AI development, while others warn of risks associated with unrestricted use. This release by Thinking Machines fits into this evolving landscape, representing a push toward more open foundational models, albeit with some caveats and restrictions.
“We believe in empowering developers and researchers with full access to our models, fostering innovation and responsible use.”
— Thinking Machines spokesperson
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.
Questions About Data Use and Licensing Restrictions
While the weights are openly available, reports suggest that Thinking Machines maintains a separate Acceptable Use Policy that could impose restrictions on surveillance, deception, and decision-making applications. The exact scope, enforceability, and legal implications of this policy remain unclear, as the full text has not been publicly verified. Additionally, the training data and pipeline have not been disclosed, raising questions about data provenance and potential biases.
Domain-Specific Small Language Models: Efficient AI for local deployment
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Industry Adoption and Oversight
Expect further analysis and independent benchmarking of Inkling and its variants, alongside scrutiny of the licensing and use policies. Organizations interested in deploying the model will need to evaluate the legal and ethical implications, especially regarding the separate use restrictions. Industry observers anticipate that this release could trigger more open models and influence future licensing practices, but also prompt discussions about safety and responsible use in open AI development.
Large Language Models: The Hard Parts: Open Source AI Solutions for Common Pitfalls
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What makes Inkling different from other foundation models?
Inkling is notable for its full open weights release under Apache 2.0, supporting multimodal input and a very large context window, with a focus on transparency and ownership, unlike most proprietary models.
Are there restrictions on how I can use Inkling?
While the weights are openly available, reports indicate that Thinking Machines has a separate Acceptable Use Policy restricting certain applications, such as surveillance and deception. Verify the policy before use.
Does the open release include training data or pipeline details?
No, the training data and full training pipeline have not been published, which limits understanding of potential biases or data provenance issues.
Will this influence future AI model releases?
Yes, this move could encourage other organizations to release more open models, but it also raises questions about safety, licensing, and data governance that may shape future standards.
Source: ThorstenMeyerAI.com
