Inkling AI

The unofficial guide to Thinking Machines' 975B open model

inklingai.site

Key Topics

Inkling 975B model 1M context window open weights multimodal model Apache 2.0

Project Review

FAQ 3

Quick Verdict

Inkling is an open-weights, multimodal model the website describes as a 975B-parameter model with a 1M-token context window. Someone might care because the project appears to be published under an Apache 2.0 license and the site includes a guide that discusses performance benchmarks and how to run the model.

Key Features

  • The website describes Inkling as a multimodal model, which indicates it is intended to handle more than one input modality (the website does not clearly state the exact modalities or file formats supported).
  • It is presented as an open-weights release (Apache 2.0 license), which suggests you can inspect and redistribute the model weights under that license.
  • The headline technical claims are a 975 billion parameter size and a 1,000,000-token (1M) context window; the site also includes a guide that “discusses performance benchmarks and instructions on how to run the model.”
  • Privacy and legal notes on the site include a privacy policy (last updated July 18, 2026) that mentions anonymous analytics and terms of service that describe the site as informational/entertainment only.

Separate confirmed facts from interpretation: the site text confirms the parameter count, context window, open-weights claim, license name, and that a guide exists; it does not provide full technical benchmarks, deployment steps, or system requirements in the material provided here.

Who this is for

Best fit: engineers, researchers, and advanced hobbyists who want an openly licensed, large-context multimodal model to experiment with locally or in research settings. The Apache 2.0 license is a trust signal for reuse and redistribution.

Verify before committing: the website does not clearly state pricing, hosting or API options, hardware requirements, detailed benchmark numbers, input/output formats, or data-retention practices. Confirm those specifics on the official site before planning production use.

FAQ

Q: What license governs Inkling?
A: The website describes the release as Apache 2.0 licensed; verify the license text on the official release to confirm scope and exceptions.

Q: Can I run Inkling locally and what hardware do I need?
A: The site includes a guide with instructions on how to run the model, but it does not clearly state detailed hardware or system requirements in the provided material—verify recommended hardware and memory needs on the official documentation.

Q: How does the project handle user data and analytics?
A: The privacy policy (last updated July 18, 2026) states the site may collect anonymous usage data via analytics; it does not clearly state data retention, training-use of inputs, or security safeguards, so confirm those policies on the website before sending sensitive data.

Frequently Asked Questions

What license governs Inkling?

The website describes the release as Apache 2.0 licensed; verify the license text on the official release to confirm scope and exceptions.

Can I run Inkling locally and what hardware do I need?

The site includes a guide with instructions on how to run the model, but it does not clearly state detailed hardware or system requirements in the provided material—verify recommended hardware and memory needs on the official documentation.

How does the project handle user data and analytics?

The privacy policy (last updated July 18, 2026) states the site may collect anonymous usage data via analytics; it does not clearly state data retention, training-use of inputs, or security safeguards, so confirm those policies on the website before sending sensitive data.

Inkling (975B) — Open-weights Multimodal Model with 1M Context Window

Independent profile of Inkling: a 975B-parameter, multimodal, open-weights model the site describes with a 1M-token context window and Apache 2.0 license. Verify deployment, pricing, and data policies on the official site.

Editorial Notice

Reviewed by the Launchtory Editorial team ·

This is an independent third-party profile of Inkling AI and is not officially affiliated with the project.

We base profiles on the official website and public product information, then verify key details against the source. AI assistance is used for drafting, with editorial review of structure and claims. Check pricing and availability on the official site before relying on them.

How we review projects

Outbound links may include a referral parameter for attribution.

From the collection · Platform-maintained information, with AI-assisted analysis where available.

Similar projects

Alternatives and adjacent projects worth comparing.

Gradio 5.0

the easiest way to build AI web apps

gradio.app

Osaurus

Open source agents that run 100% locally on your Mac

osaurus.ai

DeepClaude

DeepSeek R1 + Claude = Powerful AI Synergy

deepclaude.com

Jev Model

Fast, structured decisions with probabilities for software teams.

thejevai.com

OpenPeer AI Pre-Launch

Safe, Democratized AI for the Masses

openpeer.me

Cowork.ink

Fast, lightweight AI agent platform — open-source alternative to OpenClaw for $20/mo

Cowork.ink

Claude 3.7 Sonnet and Claude Code

Claude 3.7 Sonnet and Claude Code AI that codes and reasons.

claude.ai

Unsloth

Finetune LLMs 2x faster, 80% less memory

unsloth.ai

DeepSeek Japanese

DeepSeek in Japanese language

deepseekai.jp

DeepSeek Deutsch

https://deepseek-deutsch.de/

deepseek-deutsch.de

Useful tools

Simple tools related to this project's topics and use cases.

More

UUID Generator

Generate universally unique identifiers for development, testing, and data workflows.

Kingdom Name Generator

Generate kingdom names for fantasy maps, royal courts, tabletop campaigns, and fiction.

Orc Name Generator

Generate orc names with clan or nomad bynames for fantasy characters.

Band Name Generator

Generate distinctive band names in indie, electronic, or rock styles for music projects and fictional groups.