Inkling vs. DeepSeek-VL

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.

DeepSeek-VL is a groundbreaking open-source model that merges vision and language capabilities, specifically designed for practical use in everyday settings. Our approach is based on three core principles: first, we emphasize the collection of a wide and scalable dataset that captures a variety of real-life situations, including web screenshots, PDFs, OCR outputs, charts, and knowledge-based data, to provide a comprehensive understanding of practical environments. Second, we create a taxonomy derived from genuine user scenarios and assemble a related instruction tuning dataset, which is aimed at boosting the model's performance. This fine-tuning process greatly enhances user satisfaction and effectiveness in real-world scenarios. Furthermore, to optimize efficiency while fulfilling the demands of common use cases, DeepSeek-VL includes a hybrid vision encoder that skillfully processes high-resolution images (1024 x 1024) without leading to excessive computational expenses. This thoughtful design not only improves overall performance but also broadens accessibility for a diverse group of users and applications, paving the way for innovative solutions in various fields. Ultimately, DeepSeek-VL represents a significant step towards bridging the gap between visual understanding and language processing.