Inkling vs. Altar-1

Inkling is an open-weights multimodal AI model from Thinking Machines built to support customization, agentic workflows, coding, reasoning, vision, audio, and enterprise AI use cases. The model is a Mixture-of-Experts transformer with 975 billion total parameters, 41 billion active parameters, 256 routed experts per MoE layer, and six routed experts active per token. It supports context windows up to 1 million tokens and was pretrained on 45 trillion tokens across text, images, audio, and video. Inkling is designed as a broad foundation model rather than a narrowly optimized benchmark model, giving it balanced capabilities across reasoning, coding, factuality, instruction following, vision, audio, tool use, and safety. Its controllable thinking effort lets developers adjust how much computation and generated reasoning the model uses, helping teams balance quality, latency, and cost for different production needs. The model can run agentic coding tasks, use tools, create web apps, generate polished multi-page artifacts, reason over long contexts, and work through iterative refinement loops. For multimodal tasks, Inkling can process images, answer questions about visual content, transcribe and reason over audio, follow spoken instructions, and combine visual reasoning with code-based tools such as Python. Thinking Machines trained Inkling for calibration, instruction following, factual reliability, refusal behavior, and safety across multiple modalities, including evaluations for dangerous capabilities and human-AI threat vectors. Inkling is available on Tinker for fine-tuning, with 64K and 256K context options, an Inkling Playground for testing, cookbook recipes, and support for multimodal post-training workflows. Its full weights are available on Hugging Face, and deployment support is available through APIs and infrastructure partners such as TogetherAI, Fireworks, Modal, Databricks, Baseten, SGLang, vLLM, llama.cpp, and transformers.

The Aikido Altar is a state-of-the-art open-weight security framework designed to provide organizations with sophisticated defensive security intelligence specifically adapted to their infrastructure needs. This system is particularly effective in environments that demand sovereign security, safeguarding sensitive materials such as source code, architectural blueprints, vulnerability reports, and various confidential data within the organization's perimeter, preventing any sharing with outside inference services. Utilizing the GLM-5.3 architecture, Altar incorporates methods like quantization and expert pruning that reduce the model size from a substantial 1.51 TB in full precision to a more manageable 328 GB, while still preserving most of the original model's reasoning and security capabilities. The architecture maintains 168 of the originally 256 routed experts in each backbone expert layer and employs a W4A16 representation, improving its utility for security tasks that necessitate managing large and evolving context windows. Expert selection was calibrated using internal pentesting data alongside multilingual resources, ensuring that no client data was involved, thus maintaining the integrity of cybersecurity, programming, and language processing capabilities. This groundbreaking strategy not only simplifies the deployment process but also enhances the organization's defense against new and evolving threats, ensuring robust protection in an increasingly complex digital landscape. Furthermore, the ongoing development of additional features aims to further bolster the system’s resilience and adaptability to future security challenges.