📊 Full opportunity report: What Thinking Machines’ Inkling Is Signaling About AI’s Future on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines has released Inkling, a large open-weight AI model openly acknowledged as not the strongest available. The release highlights transparency about model capabilities and licensing, signaling a shift toward open, honest AI development.
Thinking Machines has officially released Inkling, a 975-billion-parameter open-weight AI model, marking a notable shift in the industry’s approach to transparency and openness. The company explicitly states that Inkling is not the strongest model available, signaling a different emphasis from typical industry launches that often highlight performance above all.
Inkling is a Mixture-of-Experts transformer supporting multimodal input — text, images, and audio — with a 1-million-token context window, pretrained on 45 trillion tokens. The full model weights are available on Hugging Face under the Apache 2.0 license, allowing download, modification, and commercial use. This contrasts with many recent releases that restrict access through closed APIs or proprietary licenses.
Despite its openness, the company has indicated that Inkling is not the top-performing model, with benchmarks showing it behind some closed models in certain tasks. The release was accompanied by detailed training information, including the use of synthetic data from open models like Kimi K2.5, and a candid acknowledgment of the model’s current limitations.
Additionally, reports suggest that Thinking Machines enforces a separate Model Acceptable Use Policy (AUP), which may restrict surveillance, deception, and automated decision-making, raising questions about the true openness of the model’s use and licensing framework.
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 Transparent Open-Weight Model Release
The release of Inkling underscores a potential shift toward greater transparency and openness in AI development. By openly sharing the full weights under a permissive license and acknowledging its limitations, Thinking Machines challenges the industry norm of proprietary, closed models.
This move could influence how organizations and developers approach AI ownership, favoring models that can be inspected, fine-tuned, and deployed independently. However, the potential layered restrictions via the separate AUP introduce questions about the true scope of openness and how enforceable these policies are in practice.
Overall, Inkling’s release may encourage a more honest conversation about AI capabilities, licensing, and responsible use, fostering a more open ecosystem but also highlighting ongoing tensions between openness and control.
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 Open-Weight AI Model Releases
Over the past few years, the AI industry has seen a mix of open and closed model releases, with many companies favoring proprietary approaches to protect intellectual property and commercial interests. Recent high-profile launches often involve closed APIs, with full model weights kept private to prevent misuse and maintain competitive advantage.
Thinking Machines, founded by former OpenAI CTO and staffed with experts involved in ChatGPT development, has now taken a different stance by releasing Inkling’s full weights openly, emphasizing transparency and user control. This approach echoes broader discussions within the AI community about balancing innovation, safety, and openness.
The release comes amid ongoing debates about model safety, licensing, and the potential for misuse, with some industry voices calling for more transparent practices to foster trust and collaboration.
“We believe in transparency and giving users the freedom to inspect, modify, and deploy our models under open licenses.”
— Thinking Machines spokesperson
AI Engineering: Building Applications with Foundation Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Inkling’s Licensing and Use Restrictions
It remains unclear how the separate Model Acceptable Use Policy (AUP) will be enforced and whether it imposes significant restrictions beyond the Apache 2.0 license. The precise scope of permissible modifications, commercial use, and restrictions on surveillance or automated decision-making is still to be verified through the official policy documentation.
Additionally, the long-term performance of Inkling compared to state-of-the-art models in various applications and domains is still being evaluated, with full benchmarking pending.
Mastering Small Language Models: A Practical Guide to Building Lightweight NLP Systems with Python, Transformers, and Quantization Techniques
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Industry Adoption and Evaluation
Further independent benchmarking and real-world testing will clarify Inkling’s capabilities and limitations. Industry observers will likely scrutinize the enforceability and scope of the AUP before organizations consider deploying the model in sensitive domains.
Thinking Machines plans to release additional smaller variants and continue refining the model, with ongoing updates on performance and licensing details expected in the coming months. The broader AI community will watch to see if this open approach influences future model releases and licensing practices.
AI model licensing and management
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What makes Inkling different from other large language models?
Inkling is openly available under the Apache 2.0 license, with full weights released publicly, allowing inspection, modification, and deployment without proprietary restrictions. It also supports multimodal input and emphasizes transparency about its capabilities and limitations.
Does open access mean Inkling is the best model available?
No, according to the developers, Inkling is not the strongest model on the market. Benchmarks show it lags behind some closed models in certain tasks, but its openness is its defining feature.
What are the potential risks of using an open-weight model like Inkling?
While open weights allow for customization and transparency, they can also be misused for malicious purposes such as surveillance or deception if not properly governed by usage policies. Enforcement of restrictions remains a concern.
Will Thinking Machines release more models under similar terms?
The company has indicated ongoing development and plans for additional variants, but details about future licensing and openness levels are not yet confirmed.
Source: ThorstenMeyerAI.com
