📊 Full opportunity report: What Does Thinking Machines’ Inkling Tell Us About AI’s Next Step? on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines launched Inkling, a 975-billion-parameter open-weight AI model, emphasizing transparency and honest benchmarking. The model’s open weights are available on Hugging Face under Apache 2.0, marking a significant step in open AI development.
Thinking Machines has officially released Inkling, its latest foundation model, making the full weights available on Hugging Face under the Apache 2.0 license. This marks a notable moment in AI development, emphasizing transparency and ownership over proprietary models, and directly addresses ongoing debates about open-source AI’s role and restrictions.
Inkling is a 975-billion-parameter, multimodal transformer designed to process text, images, and audio jointly, supporting a one-million-token context window. It was pretrained on 45 trillion tokens across various modalities, with a focus on transparency and open access. The model’s weights are publicly available, allowing users to download, modify, and deploy independently, a departure from typical proprietary AI models.
Thinking Machines explicitly stated that Inkling is not the strongest model available today, but its open-access approach is significant. The model was trained using a hybrid optimizer on NVIDIA hardware, with reinforcement learning improving reasoning performance. A smaller version, Inkling-Small, is also in testing, showing promising benchmark results. The release includes a detailed license and hints at a separate Model Acceptable Use Policy, which may impose restrictions beyond the Apache license, raising questions about the scope of openness.
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 Model Release
The release of Inkling under an open license signifies a shift towards greater transparency and ownership in AI development, enabling organizations to fine-tune, inspect, and deploy the model independently. This challenges the traditional model of renting AI via APIs and raises important questions about data privacy, licensing restrictions, and responsible use. The transparency around the model’s capabilities and limitations provides a new benchmark for evaluating AI safety and performance, particularly in safety-critical domains.
However, the potential layered restrictions through a separate Acceptable Use Policy could complicate the open-source narrative, especially if they limit surveillance, deception, or automated decision-making. This development may influence industry standards, regulatory policies, and the future of open AI research, emphasizing the importance of clear licensing and usage terms.
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.
Background on Open-Weight AI Models and Industry Norms
Historically, AI models like GPT-3 and others have been released as closed proprietary systems, with only API access provided to users. Open-source efforts, such as those on Hugging Face, have offered weights under licenses like Apache 2.0, but often with limited transparency regarding training data and pipelines. Recent industry debates focus on the trade-offs between openness, safety, and commercial viability.
Thinking Machines, founded by former OpenAI CTO, has a reputation for building advanced models. Its decision to release Inkling’s full weights openly, while openly acknowledging it isn’t the strongest, marks a deliberate move towards transparency. The company’s approach contrasts with recent industry trends, where many firms prefer controlled API access or closed weights to retain commercial control.
“Inkling is not the strongest model today, but its open access under Apache 2.0 provides maximum flexibility for users.”
— Thinking Machines spokesperson
The Practical Guide to Large Language Models: Hands-On AI Applications with Hugging Face Transformers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Inkling’s Use Restrictions
It remains unclear how the separate Model Acceptable Use Policy will be enforced and what specific restrictions it imposes beyond the Apache 2.0 license. The extent to which these restrictions could limit commercial or research applications is still unknown, as the policy has not been publicly verified.
Additionally, the long-term implications of layered licensing and restrictions on open-source models are still developing, and industry consensus on best practices remains unsettled.
AI Engineering: Building Applications with Foundation Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Future Developments and Industry Impact of Inkling
Next steps include independent benchmarking of Inkling’s performance, verification of the Acceptable Use Policy, and broader industry responses to open-weight releases. Companies and researchers will likely test the model’s capabilities in real-world applications, assess safety and compliance, and evaluate how layered restrictions influence adoption. Further updates from Thinking Machines on testing and licensing clarifications are expected in the coming months.
Laplink PCmover Migration Software – Initial Pay-Per-Use License Fee – Monthly invoicing for additional uses – $29.95/license with Super Speed USB 3.0 cable – Business Technician, 10 Licenses
Flexible Pay-Per-Use Structure: Laplink's Technician licensing bills only for completed transfers. One license covers unlimited transfer attempts from…
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 notable for its full open weights under Apache 2.0, allowing free download, modification, and deployment, unlike most proprietary models. It also supports multimodal input and has a large context window, making it versatile for various applications.
Does open weights mean the model is fully open source?
No. While the weights are open under Apache 2.0, the training data, training pipeline, and any layered Acceptable Use Policy may impose restrictions, meaning it is not fully open source in the traditional sense.
What are the potential risks of layered licensing restrictions?
Layered restrictions could limit the model’s use in certain domains, create legal uncertainties, and undermine transparency if enforcement is inconsistent or opaque.
Why does this release matter for the AI industry?
It signals a shift toward greater transparency and ownership options in AI, challenging the dominance of closed models and potentially influencing industry standards and regulatory policies.
Source: ThorstenMeyerAI.com
![Governments, companies, nonprofits should invest in free, open source AI [pdf]](https://auto-blogging.com/wp-content/uploads/2026/07/governments-companies-nonprofits-should-invest-in-free-open-source-ai-pdf-featured-1-560x300.jpg)
