Ling 3.0 Tiny vs. Inkling

Ling 3.0 Tiny is an advanced reasoning model with open weights, consisting of 7.9 billion parameters in total and 1.3 billion that are active, while boasting a remarkable context window of 262,000 tokens. Utilizing a mixture-of-experts architecture, it expands the open-weights Pareto frontier in intelligence relative to its active parameters, all while maintaining a compact size suitable for deployment in various settings. With a score of 25 on the Artificial Analysis Intelligence Index, it rivals gpt-oss-120b, which has a score of 24, even though it uses 15 times fewer total parameters and 4 times fewer active parameters. This exceptional efficiency in parameters comes with a cost, as it demands a hefty 213 million output tokens to finalize the Intelligence Index evaluation. Moreover, Ling 3.0 Tiny shows significant progress in mitigating hallucination rates when compared to Ling-mini-2.0; it boosts its AA-Omniscience score by an impressive 59 points while maintaining consistent accuracy. Rather than resorting to random guesses in uncertain scenarios, the model opted to attempt only 37% of the posed questions during assessment, which resulted in a drastically lowered hallucination rate of 30%, a substantial improvement from the previous generation's staggering 96%. This strategic decision not only underscores the model's enhanced reasoning abilities but also emphasizes its potential for practical applications in the real world. Overall, Ling 3.0 Tiny exemplifies a significant step forward in the development of efficient and reliable AI models.

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.