Thinking Machines launched Inkling on July 15, 2026, an open-weight multimodal AI foundation model designed for enterprise and developer fine-tuning rather than frontier competition. The model is a Mixture-of-Experts transformer with 975 billion total parameters and 41 billion active parameters, supporting a context window of up to one million tokens. The company positions Inkling as a flexible base for customization, emphasizing controllable thinking effort and native multimodal reasoning across text, images, and audio. Full model weights are available on Hugging Face, with fine-tuning accessible via the company's Tinker platform. Thinking Machines explicitly states the model does not claim state-of-the-art status but focuses on breadth of capability, cost efficiency, and safety calibration for enterprise deployment.
Thinking Machines Releases Technical Specifications for Inkling Model
Inkling is pretrained on 45 trillion tokens spanning text, images, audio, and video. The model offers native multimodal reasoning across all three input types, a capability that distinguishes it from most open-weight alternatives, which typically lack native audio support. Developers can tune how many tokens the model uses to solve a problem, enabling cost and latency savings. In testing, Inkling matched Nemotron 3 Ultra on Terminal Bench 2.1 at roughly one-third the token cost.
Thinking Machines also previewed Inkling-Small, a lighter variant with 276 billion total parameters and 12 billion active parameters. Inkling-Small matches or exceeds the larger model on several benchmarks, offering a lower-cost option for synthesis and grading workloads.
Inkling Achieves Competitive Benchmark Results Against Closed-Weight Models
Benchmark results show competitive but not leading performance compared to closed-weight models such as Claude Fable 5 and GPT-5.6 Sol on reasoning and agentic tasks. The release emphasizes strong performance across coding, instruction following, factuality, vision, and audio.
On ForecastBench, Inkling performs on par with leading closed models including Gemini 3.1 Pro and Grok 4.3. On FORTRESS, a benchmark evaluating refusal of harmful requests while avoiding over-refusal of benign analogs, Inkling scored 78% on adversarial prompts against 77.6% for Nemotron 3 Ultra and 65.6% for Kimi K2.6.
Thinking Machines Implements Safety and Epistemic Calibration in Inkling Training
Thinking Machines trained Inkling using reinforcement learning against proper scoring rules on a large corpus of resolved real-world forecasting questions, producing a model calibrated to express appropriate uncertainty rather than confidently hallucinating. The training pipeline incorporated dual automated graders — a rubric grader and a claims grader with agentic web search — to simultaneously improve helpfulness and reduce factual errors.
Thinking Machines Makes Inkling Available Through Multiple Deployment Partners
Both Inkling and Inkling-Small are available through Tinker. Deployment partnerships span TogetherAI, Fireworks, Databricks, Hugging Face, and others.
FAQ
What did Thinking Machines launch on July 15, 2026?
Thinking Machines launched Inkling, an open-weight multimodal AI foundation model with 975 billion total parameters and 41 billion active parameters, designed for enterprise and developer fine-tuning.
How does Inkling compare to closed-weight models on benchmarks?
Inkling shows competitive but not leading performance compared to closed-weight models such as Claude Fable 5 and GPT-5.6 Sol on reasoning tasks. On ForecastBench, it performs on par with Gemini 3.1 Pro and Grok 4.3. On FORTRESS, it scored 78% on adversarial prompts.
Where can developers access Inkling for fine-tuning?
Full model weights are available on Hugging Face, and fine-tuning is accessible via the company's Tinker platform. Deployment partnerships include TogetherAI, Fireworks, Databricks, and Hugging Face.
