📊 Full opportunity report: The Rise Of Inkling: A New Era For Thinking Machines on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Thinking Machines has released Inkling, a 975-billion-parameter multimodal AI model, on Hugging Face. The model can process text, images, and audio within a large context window but requires substantial hardware. Its benchmarks and licensing details remain unverified.
Thinking Machines has released Inkling, a 975-billion-parameter multimodal model, on Hugging Face, offering new capabilities for processing text, images, and audio within a claimed one-million-token context window. This release marks a significant step in open AI models, though its hardware demands and lack of independent benchmarks limit immediate widespread use.
Inkling is described as a decoder-only Mixture-of-Experts model with 975 billion total parameters and 41 billion active during processing. It was trained on 45 trillion tokens across multiple data types, including text, images, audio, and video, though these figures have not been independently verified. The architecture employs 256 experts, with dynamic routing and attention mechanisms designed to optimize multimodal reasoning.
Hardware requirements are substantial: Hugging Face states that the BF16 checkpoint needs approximately 2 TB of VRAM, while the NVFP4 version requires about 600 GB, making it impractical for most consumer systems. The release includes support for popular inference frameworks, and developers can access the model via hosted services, but detailed licensing and performance benchmarks are not yet available. For more insights, see the original analysis.
Implications of Inkling’s Large-Scale Multimodal Capabilities
The release of Inkling signifies a major advancement in multimodal AI, potentially enabling more integrated and complex reasoning across different data types. Its open availability could accelerate research in scientific, media-analysis, and enterprise applications, though hardware constraints limit immediate adoption for many users. The lack of independent benchmarks and licensing clarity raises questions about its safety, reliability, and commercial viability in the near term.
GIGABYTE Radeon™ AI PRO R9700 AI TOP 32G Graphics Card, Turbo Fan Cooling System, 32GB GDDR6, GV-R9700AI TOP-32GD Video Card
Powered by Radeon AI PRO R9700 – Supercharge you workflow with the cutting-edge RDNA 4 Architecture and 2nd-gen…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on Large Multimodal AI Models and Recent Developments
Prior to Inkling, most large multimodal models have been proprietary or limited in scale, often requiring specialized hardware. The recent trend has been toward increasing model size and multimodal integration, but practical deployment has remained challenging due to hardware demands. The open release of Inkling by Thinking Machines builds on this trend, offering a model of unprecedented size and multimodal scope, though with limited independent evaluation or detailed licensing information.
“This model is huge.”
— Hugging Face spokesperson
Yahboom Raspberry Pi 5 ROS2 Robot Car 360°Movement, AI Vision & Tracking, Integrated Multimodal Large AI Model OpenRouter, AI Voice Interaction (Superior Without RPi5)
【Powerful control system】RaspberryPi 5 has made breakthroughs in processor speed,multimedia performance,memory and connection.Based on the RaspberryPi 5 main…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unverified Benchmark Results and Licensing Details
Independent evaluations of Inkling’s performance, safety, and robustness are not yet available. The licensing terms remain unspecified, and the impact of the large context window on real-world workloads is still unclear. How well the model performs on video or in consumer hardware settings also remains to be seen.
NIMO AI NAS, Agentic Computer Mini PC and AI Server, AMD Ryzen 7 PRO 8845HS
Next-Gen AI & LLM Local Deployment: Powered by the 8845HS processor and RTX 5060 GPU, this NAS provides…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Expected Developer Testing and Benchmarking Efforts
Early testing by researchers and developers will focus on latency, accuracy, and hardware compatibility, especially using the supported inference frameworks. Benchmark results, safety assessments, and licensing clarifications are anticipated in the coming months, which will determine the model’s practical viability and adoption scope.
AI in Embedded Systems: Types, Techniques, Machine Learning, Model Training vs. On-device Inference, Algorithms, Frameworks and Tools.
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Inkling?
Inkling is a large, open multimodal AI model from Thinking Machines, capable of processing text, images, and audio, with 975 billion parameters and a large context window.
Can I run Inkling on my personal computer?
Unlikely. The model’s hardware requirements are extremely high, with estimates of 2 TB of VRAM for full deployment, making it suitable mainly for specialized hardware or hosted services.
Does Inkling support video processing?
The architecture accepts image inputs with a temporal dimension, but native video performance has not been evaluated or confirmed.
What are the licensing terms for Inkling?
The release describes Inkling as an open model but does not specify licensing details, usage restrictions, or whether the training data and code are publicly available.
When will independent benchmarks and safety evaluations be available?
There is no confirmed timeline, but early testing and evaluations by the developer community are expected to provide insights in the coming months.
Source: ThorstenMeyerAI.com
