Inkling (Thinking Machines Lab)

What it can do

• Generate reasoned text responses over long documents

  Text prompt with up to 1M tokens of context → Reasoned text response

• Analyze and answer questions about images

  Image plus text query → Text description or answer

• Process and reason over audio input

  Audio file plus text query → Text transcription, summary, or answer

• Reason jointly across text, images, and audio in a single prompt

  Mixed multimodal input (text, images, audio) → Unified reasoned text response

• Fine-tune the base model on custom data via Tinker

  Custom training dataset → Customized fine-tuned model weights

• Run locally or self-host as an open-weights model

  Downloaded Apache 2.0 model weights → Deployed inference-ready model

• Summarize large context windows of documents or conversations

  Long-form text up to 1M tokens → Concise summary text

Why it made the leaderboard

Inkling is a 975B-param (41B active) multimodal MoE released under Apache 2.0 with 1M-token context and same-day fine-tuning on Tinker, giving builders a permissively licensed U.S. open base model to customize rather than a locked-down API. Its novel architecture (relative positional bias, scaled short-conv layers, dual shared-expert MoE) makes it a distinct alternative to Llama, DeepSeek, and GLM/Kimi for teams needing open weights they can adapt.

Media

Comments (0)

No comments yet