# Inkling: Open-Weights Multimodal Model Announced

Customizable AI model Inkling now available for developers worldwide.

- **Source:** [Hacker News (news.ycombinator.com)](https://brifing.si/source/13c923cd-6d99-442b-9b8b-1181f4d8cd91)
- **Published:** 2026-07-15
- **Topic:** [Data & AI](https://brifing.si/c/data-ai)
- **Original article:** <https://thinkingmachines.ai/news/introducing-inkling/>
- **On Brifing:** https://brifing.si/s/ba7ace09-0f2d-4516-9448-22b1b876c8b6

## Next-Gen AI Model

Inkling launches as an open-weights AI model with 975B parameters. It offers multimodal capabilities processing text, images and audio. Full access provided for developer customization.

## Technical Specifications

Trained on 45 trillion tokens with 1M token context window. Uses Mixture-of-Experts architecture. Lighter Inkling-Small version also available with 12B parameters.

## Multimodal Capabilities

Inkling natively processes visual and auditory data alongside text. Balances different thinking efforts for cost-performance optimization. Designed for broad application domains.

## Tinker Platform Integration

Model ready for fine-tuning on Tinker platform. Developer-facing Inkling Playground interface created. Capable of writing and running its own fine-tuning jobs.

## General Performance

Inkling delivers balanced performance across domains rather than narrow specialization. Shows competitive results in agentic coding, tool use and multimodal evaluations.

## Web Development Capabilities

Ranks high on Design Arena's Agentic Web Dev leaderboard. Can create functional web apps in single pass. Stands out among open-weights models in human evaluations.

## Collaborative Refinement

Inkling works through long refinement cycles. Developed an online game through 40 feedback iterations. Capable of producing multi-page cohesive designs.

## Controllable Thinking Effort

Model's thinking depth can be adjusted by users. Test-time scaling and problem-solving are core features. Optimizable for different workloads.

## Safety and Reliability

Model underwent comprehensive safety and trustworthiness evaluations. Carries qualities necessary for real-world applications. Adaptable to various usage scenarios.

## Customization Potential

Open-weights structure enables extensive customization options. Designed as foundational model adaptable to different workflows. Users can shape it according to their needs.

## Future Roadmap

Inkling positioned as first in a family of models. New versions in different sizes planned. Goal to expand customization possibilities.

## Contribution to AI Ecosystem

Open-weights model access diversifies AI ecosystem. Offers developers more control and flexibility. Inkling supports mission to extend human will through AI.

## Sources covering this story

- https://thinkingmachines.ai/news/introducing-inkling/ — 

---

_Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._