📊 Full opportunity report: What The Inkling From Thinking Machines Tells Us About AI Innovation on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines has released Inkling, a large, open-weight AI model, openly acknowledging it is not the strongest available. This move emphasizes transparency and ownership in AI development, but raises questions about licensing and use restrictions.
Thinking Machines has publicly released the full weights of its new AI model, Inkling, under the open-source Apache 2.0 license. This marks a notable departure from typical industry practice, emphasizing transparency and ownership, and directly addressing ongoing debates about model control and licensing.
Inkling 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 multiple modalities, with a hybrid optimizer and over 30 million reinforcement learning rollouts. The full weights are now available on Hugging Face, with day-zero support in several open-source frameworks, enabling broad access and modification.
Despite its openness, the company clarifies that the weights are not open source in the traditional sense, as the training data and pipeline remain proprietary. Additionally, Thinking Machines reportedly enforces a separate Model Acceptable Use Policy that restricts surveillance, deception, and automated decisions affecting individuals, raising questions about the scope of its openness.
Industry observers note that Inkling is not the strongest model available, but its open weights and transparent approach mark a strategic shift. Benchmarks show it excels in safety and speech tasks but is mid-tier in some language understanding metrics, reflecting a balanced but not top-of-the-line performance profile.
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 signals a shift toward greater transparency and ownership in AI development, allowing organizations to download, fine-tune, and deploy models independently. It challenges the industry norm of closed models and highlights ongoing debates about licensing, data privacy, and responsible use. The approach may influence future releases, encouraging a more open and accountable AI ecosystem, but also raises questions about enforceability of usage restrictions embedded in policies.
Yahboom K230 AI Development Board 1.6GHz High-performance chip/2.4-inch Display/Open Source Robot Maker Python, Supports AI Visual Recognition CanMV Sensor (with Heightened Bracket)
【Flagship performance, extremely fast response】Equipped with a 1.6GHz main frequency chip, the KPU computing power is 13.7 times…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Industry Trends Toward Transparency and Open Models
Over recent years, the AI industry has seen a growing divide between closed, proprietary models and open-source initiatives. Major players often withhold weights or restrict access, citing safety and competitive concerns. The recent release of Inkling by Thinking Machines, founded by former OpenAI executives, reflects a strategic move toward openness, inspired by broader calls for transparency and democratization of AI technology. This development comes amid ongoing debates about the balance between openness, safety, and commercial interests.
“We believe in giving users full control over the models they choose to deploy, which is why we released Inkling under Apache 2.0 and included detailed licensing information.”
— Thinking Machines spokesperson
Engineering with Small Language Models: Efficient AI Design, Training, and Deployment for Developers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Licensing and Use Restrictions
It remains unclear how enforceable the separate Model Acceptable Use Policy is, and whether it effectively limits how the open weights can be used. The full scope of restrictions, data provenance, and compliance mechanisms has not been independently verified, raising questions about practical control over the model’s deployment and modifications.
HIWONDER Humanoid Robot with ChatGPT Multimodal AI Models AI Embodied Intelligent Vision Scene Voice Understanding 20DOF Educational Robotic Kit Python Programming, TonyPi Pro & RaspberryPi 5 16GB
Al-Driven & Raspberry Pi Powered. TonyPi is a high-performance AI vision robot designed for AI education applications. It…
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 researchers and industry stakeholders on the enforceability of the use policy and the model’s safety features. Additional benchmarks and real-world testing will clarify Inkling’s capabilities and limitations. Meanwhile, other organizations may follow suit, releasing open weights with layered restrictions, prompting ongoing debate about transparency versus control in AI development.
AI model licensing and safety tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Why is the release of Inkling significant?
The full weights are now publicly available under an open license, enabling independent use, modification, and deployment, marking a shift toward transparency in AI development.
What makes Inkling different from other models?
It is a large, multimodal model with open weights, trained with transparent methods, and accompanied by a layered use policy that restricts certain applications.
Does open weights mean unrestricted use?
Not necessarily. While the weights are openly available, the company reportedly enforces a separate use policy that may restrict some applications, though the enforceability of these restrictions remains uncertain.
What are the risks of releasing open weights?
Open weights can be misused for malicious purposes, misinformation, or surveillance, especially if layered restrictions are not enforceable or transparent.
What will happen next in AI openness?
Expect more organizations to release models with layered licensing and use policies, prompting ongoing discussions about balancing transparency, safety, and control.
Source: ThorstenMeyerAI.com
