Inkling-Small vs. Inkling 2 Small

Inkling-Small is an efficient multimodal AI model built to deliver strong reasoning and coding performance at a fraction of Inkling’s size. It is a Mixture-of-Experts transformer with 276 billion total parameters and 12 billion active parameters. The model was trained on NVIDIA GB300 NVL72 systems and is designed to combine high capability with more efficient inference. Inkling-Small supports native reasoning across text, images, and audio, allowing it to work across multimodal tasks without relying on separate encoders. Its context window supports up to one million tokens, making it useful for long-form reasoning, large-scale code understanding, document analysis, and agentic workflows. Users can adjust reasoning effort from minimal to extra high depending on whether they need faster responses or deeper computation. The model’s training process includes improved pre-training data, post-training with on-policy distillation from Inkling, and extended agentic coding reinforcement learning. These techniques helped Inkling-Small outperform its larger counterpart on reasoning and coding benchmarks. The model performs well in coding and tool-use harnesses and exceeds 80% on SWE-bench Verified. Its encoder-free architecture processes audio as dMel spectrograms and images as 40-by-40-pixel patches alongside text tokens. By combining efficient MoE design, one-million-token context, adjustable reasoning effort, multimodal processing, coding strength, and tool-use performance, Inkling-Small is designed for developers and teams that need capable AI with lower active compute requirements.

Inkling 2 Small is an advanced multimodal large language model developed by Thinking Machines Lab to support autonomous coding, complex reasoning, visual understanding, and customizable AI applications. The model uses a sparse mixture-of-experts transformer architecture containing 594 billion total parameters and 17 billion active parameters, with 48 decoder layers and 384 specialized experts. Its architecture combines local and global attention mechanisms with a vision transformer that processes images alongside text within a shared representation. Inkling 2 Small supports a context window of up to one million tokens, making it suitable for analyzing extensive documents, large software repositories, and long-running agent workflows. Configurable reasoning effort enables users to adjust the computational resources devoted to individual tasks based on their complexity and performance requirements. Its agentic software engineering capabilities include generating code, debugging applications, implementing features, verifying changes, and operating development tools across extended workflows. The model can also create interactive websites, games, three-dimensional simulations, and visual applications while using image-based feedback to evaluate and improve its results. Advanced research capabilities allow autonomous agents to conduct machine learning experiments, optimize GPU kernels, evaluate training strategies, and iteratively refine technical solutions. Thinking Machines Lab trained the model using 8.4 trillion pretraining tokens and a large reinforcement learning process involving more than 17 million rollouts across coding, scientific reasoning, visual tasks, knowledge work, and tool use. Published evaluation results include 25.4% on FrontierCode 2.0, 72.1% on DeepSWE 1.1, 86.3% on SWE-Bench Pro v2 Hard, and an Artificial Analysis Intelligence Index score of 31.