Muse Ember vs. Inkling

Muse Ember is an efficient multimodal artificial intelligence model developed by Meta Superintelligence Labs for autonomous agent execution, coding assistance, complex reasoning, and large-scale business automation. The model is designed to handle the frequent, structured operations performed by AI agents, including tool calls, information retrieval, workflow routing, document extraction, and software development subtasks. It supports a context window of one million tokens and can generate up to 512,000 output tokens, allowing applications to process large repositories, extensive documents, and complex task histories. Muse Ember natively processes text, images, audio, video, and PDF content, enabling multimodal analysis without requiring separate models for each supported input type. Its visual understanding capabilities allow agents to interpret screenshots, charts, scanned documents, and interface elements while supporting computer-based automation workflows. Software engineering capabilities include reading source files, running tests, analyzing repositories, implementing targeted fixes, and completing programming tasks within autonomous coding environments. The model also supports enterprise automation workflows that involve retrieving information, evaluating records, executing tools in sequence, and updating data across connected business applications. Meta trained Muse Ember using Muse Wonder Full, creating a specialized model intended to combine strong reasoning capabilities with lower inference costs for frequently repeated tasks. Published evaluation results include 37.8% on AutomationBench-AA, 61.6% on Agents' Last Exam, 40.9% on Terminal-Bench 4.0, and 80.6% on PerceptionBench with tools. Muse Ember can be combined with larger models such as Muse Wonder and Muse Spark, allowing advanced models to handle planning and complex decisions while Ember performs high-volume execution tasks.

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.