📊 Full opportunity report: AI’s Next Big Step? Insights From Thinking Machines’ Inkling on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines has launched Inkling, a large open-weight AI model with 975 billion parameters, openly available on Hugging Face under Apache 2.0. The release emphasizes transparency and honesty about its performance, marking a significant step in open AI development.
Thinking Machines has officially released the full weights of its latest multimodal AI model, Inkling, making it available on Hugging Face under the open-source license Apache 2.0. This marks a significant development in AI transparency, as the company openly admits that Inkling is not the strongest model currently available, but emphasizes its commitment to open access and honesty about its capabilities.
The Inkling model is a Mixture-of-Experts transformer with 975 billion total parameters and 41 billion active. It supports a 1-million-token context window and was trained on 45 trillion tokens across text, images, audio, and video. Notably, it is multimodal from input to output, processing text, images, and audio without relying on vision adapters, with training data and model architecture designed for joint processing.
Released openly on Hugging Face under Apache 2.0, the model’s weights can be downloaded, modified, and deployed independently. However, the company’s separate Model Acceptable Use Policy (AUP) reportedly restricts certain uses, such as surveillance and automated decision-making, raising questions about the true openness of the release. The company’s transparency includes sharing benchmark scores from external sources, which show strengths in speech and safety benchmarks but middling performance in pure text tasks.
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 for AI Development and Openness
This release signals a shift toward greater transparency in the AI community, emphasizing open access over proprietary control. By releasing the full weights and openly acknowledging its limitations, Thinking Machines challenges the industry norm of guarded model releases. The move could influence how future models are shared, with potential impacts on innovation, regulation, and responsible AI deployment.
The Agentic AI Bible: The Complete and Up-to-Date Guide to Design, Develop, and Scale Goal-Driven, LLM-Powered Agents that Think, Execute and Evolve
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Industry Trends Toward Open-Weight Models
Over the past year, there has been increasing debate about the transparency and openness of large AI models. Companies like OpenAI and Meta have released models with varying degrees of openness, but many restrict access or impose licensing conditions. Thinking Machines’ approach with Inkling, emphasizing open weights and honest performance disclosures, aligns with a broader movement toward democratizing AI research and development. The release also follows recent government and industry calls for more transparency and safety in AI deployment.
“We believe in openness and honesty about our models’ capabilities and limitations. Our goal is to foster responsible innovation.”
— 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.
Questions About Usage Restrictions and Long-term Access
It remains unclear how the reported separate Model Acceptable Use Policy (AUP) will be enforced and whether it limits the model’s open-source nature. The exact scope of restrictions and their legal enforceability are not yet verified, raising questions about how freely the model can be used in practice.
NVD RTX PRO 6000 Blackwell Professional Workstation Edition Graphics Card for AI, Design, Simulation, Engineering – 96GB DDR7 ECC Memory – 4th Gen RT/5th Gen Tensor Core GPU – OEM Packaging
[NVIDIA Blackwell Streaming Multiprocessor] The new SM features increased processing throughput, and new neural shaders that integrate neural…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps in Model Adoption and Benchmark Validation
Expect independent researchers and developers to test Inkling’s performance across various domains, with particular focus on safety, bias, and real-world applications. Further clarification on the AUP and licensing terms will likely emerge as the model is adopted more widely, potentially influencing future open-weight releases.
Local LLM Inference Optimization: A Comprehensive Guide to Quantization, Hardware Acceleration, and Efficient Private AI Deployment
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Why is the open release of Inkling significant?
The open release allows researchers and developers to inspect, modify, and deploy the model independently, promoting transparency and responsible AI development.
What are the limitations of Inkling’s openness?
Although the weights are open under Apache 2.0, reports suggest there may be additional restrictions via a separate Acceptable Use Policy, which could limit certain applications.
How does Inkling compare to other models?
Benchmark scores show strong performance in speech and safety tasks but middling results in some text-only benchmarks. Its multimodal design is a notable feature.
What does this mean for the future of AI models?
This move could set a precedent for more open and transparent model releases, encouraging industry-wide shifts toward openness and responsible use.
Source: ThorstenMeyerAI.com
