Inkling vs. Bonsai 27B

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.

Bonsai 27B emerges as the newest flagship in the Bonsai series, representing the first-ever 27B-class model crafted for mobile device functionality. Leveraging the foundation of Qwen3.6 27B, this model significantly enhances local device capabilities with sophisticated multi-step reasoning, structured tool interactions, vision tasks, and agentic loops that ensure coherence across numerous operations. The Bonsai 27B is offered in two unique versions, with the Ternary Bonsai 27B utilizing ternary weights alongside FP16 group-wise scaling to achieve an effective weight of 1.71 bits, while maintaining a 5.9 GB footprint ideal for high-performance laptops. Alternatively, the 1-bit Bonsai 27B adopts binary weights with the same group-wise scaling approach, resulting in an effective weight of 1.125 bits and a reduced size of 3.9 GB, which aligns well with the memory limitations of devices such as the iPhone 17 Pro. Both variants operate smoothly throughout the entire language network, encompassing embeddings, attention mechanisms, MLPs, and the language model head, without the need for higher-precision solutions. Additionally, they include a compact 4-bit vision tower, which empowers on-device workflows to accurately analyze screenshots, documents, and camera inputs, thereby improving user interaction and productivity. This groundbreaking methodology illustrates Bonsai 27B's dedication to advancing the frontiers of mobile AI technology and enhancing the user experience across diverse applications.