📊 Full opportunity report: Thinking Machines’ Inkling As A Predictor Of AI’s Trajectory on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines has publicly released its large multimodal model, Inkling, under an open license, marking a notable move toward transparency in AI development. The model’s specifications and licensing are confirmed, but some details about usage restrictions remain uncertain.
Thinking Machines has made its Inkling model publicly available, releasing the full weights under the Apache 2.0 license on Hugging Face. This move is notable because it diverges from typical industry practice by openly sharing a large, multimodal foundation model without a closed API or restricted access, emphasizing transparency and ownership.
Inkling is a 975-billion-parameter mixture-of-experts transformer supporting multimodal inputs, including text, images, and audio. It was trained on 45 trillion tokens from diverse data types and features a 1-million-token context window. The model supports a fully open licensing model via Apache 2.0, allowing users to download, modify, and deploy independently. The weights are available on Hugging Face, with initial testing indicating it is not the strongest model on the market, but it demonstrates competitive safety and multimodal capabilities.
In addition, a smaller variant, Inkling-Small, with 276 billion parameters, is also previewed and shows promising benchmark results, thanks to an improved pre-training recipe. The training process involved hybrid optimization and extensive reinforcement learning, including synthetic data generated by open models like Kimi K2.5. The model’s open weights mark a significant shift towards transparency, though some restrictions on use may still apply, as reports suggest a separate Acceptable Use Policy (AUP) that could limit certain applications.
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-Source Large Models in AI
The release of Inkling under an open license signifies a potential shift in AI development, emphasizing transparency, ownership, and community-driven innovation. By openly sharing the full weights, Thinking Machines enables organizations to fine-tune, inspect, and deploy the model independently, reducing reliance on proprietary APIs and fostering broader experimentation. However, the possible existence of a separate AUP introduces questions about the scope of permissible use, especially in sensitive domains such as surveillance or automated decision-making. This move could influence how other AI labs approach model sharing and licensing in the future, balancing openness with responsible use.
AI multimodal model development tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Industry Practices and the Significance of Open Weights
In recent years, most foundation models have been released with limited access, often through APIs or with closed weights, to control usage and manage risks. Open-source models like Meta’s Llama or EleutherAI’s GPT variants have demonstrated the benefits of transparency but often come with restrictions or incomplete data disclosures. Thinking Machines’ approach with Inkling marks a departure by providing full weights immediately, aligning with a broader push for open AI development. The company’s candid acknowledgment that Inkling is not the top-performing model reflects a realistic positioning amid a competitive landscape dominated by closed models from industry giants.
Previous releases, such as Meta’s Llama 2, have also emphasized open licensing, but the distinction between open weights and open source remains critical. The potential for restrictions via separate policies, as reported, complicates the narrative and underscores ongoing industry debates about transparency versus responsible use.
“We believe in transparency and ownership. Our release allows users to own and modify the model without dependency on proprietary APIs.”
— A representative from Thinking Machines
Open source AI model training software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unclear Aspects of Inkling’s Usage Restrictions
While the weights are openly available, reports suggest that Thinking Machines maintains a separate Model Acceptable Use Policy (AUP) that could restrict certain applications, such as surveillance or automated decision-making affecting individuals’ rights. The exact scope and enforceability of this policy are not yet verified, raising questions about the true level of openness and control for end users.
GPU for machine learning
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Inkling’s Adoption and Evaluation
Further independent testing and benchmarking of Inkling will clarify its performance relative to other models. Monitoring how organizations interpret and implement the AUP will be crucial, especially in sensitive sectors. Additionally, more details about the training data and the full licensing terms are expected to emerge, shaping the model’s adoption trajectory and influence on industry standards.
AI model licensing and deployment tools
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 a 975-billion-parameter multimodal model with open weights under Apache 2.0, supporting text, images, and audio inputs. Its open licensing and full weight release distinguish it from many proprietary models.
Are there restrictions on how I can use Inkling?
While the weights are openly available, reports suggest a separate Acceptable Use Policy (AUP) may impose restrictions on certain applications. Users should review this policy before deployment.
Why is open-sourcing a model important?
Open-sourcing allows organizations to own, inspect, modify, and deploy models independently, fostering transparency, innovation, and control over AI systems.
Will Inkling outperform other models?
According to initial benchmarks, Inkling is not the top performer but shows strong safety and multimodal capabilities. Its primary value lies in openness and ownership rather than outright performance.
Source: ThorstenMeyerAI.com
