Inkling 2 Small
Thinking Machines Lab's 594B open-weight model that matches Gemini 3.8 Flash at a sixth of the cost per task.
Open-weight multimodal reasoning model with 594B total parameters, a 1M-token context, and Apache 2.0 weights for self-hosting and fine-tuning.
Inkling 2 Small model card on Hugging Face
1 / 2Overview
What is Inkling 2 Small?
Thinking Machines Lab's 594B open-weight model that matches Gemini 3.8 Flash at a sixth of the cost per task.
Inkling 2 Small is Thinking Machines Lab's second-generation model, released on November 5, 2026. It is a mixture-of-experts model with 594 billion total parameters and 17 billion active per token, accepting text and image input with text output and a context window of up to 1 million tokens. Artificial Analysis scores it 31 on its Intelligence Index, level with Google's Gemini 3.8 Flash at high effort, at $1.18 per index task versus $7.77 for Gemini 3.8 Flash. The full weights were published on Hugging Face on November 12, 2026 under the Apache 2.0 license, in NVFP4 and MXFP8 checkpoints. Hosted access is available through the lab's Tinker platform and Tinker Playground, and through API providers including NVIDIA, Fireworks, Together, Baseten, Nebius, Modal, and Cloudflare, plus a free route on OpenRouter.
Platforms and languages
Inkling 2 Small Availability
Platforms
Languages
Capabilities
Inkling 2 Small Key Features
594B MoE, 17B active
Mixture-of-experts design with 594B total parameters and 17B active per token, plus image input.
Apache 2.0 open weights
Full weights on Hugging Face in NVFP4 and MXFP8, free to download, fine-tune, and self-host.
1M-token context
Handles long documents, large codebases, and extended agent sessions in one context window.
Agent-first training
Trained with reinforcement learning on long-horizon coding and agent tasks, per the lab.
Adjustable thinking effort
Dial reasoning depth up or down to trade speed for thoroughness on hard tasks.
Low API pricing
$0.30 per million input tokens and $1.00 per million output tokens on Tinker.
Best for
Who uses Inkling 2 Small?
Developers building coding agents and long-horizon AI workflows
Run a frontier-class open model via API or self-hosted weights
Teams that need a capable model they can fine-tune and control
Download Apache 2.0 weights and adapt the model to private data
Plans and access
Inkling 2 Small Pricing
Freemium
$0.30/M input, $1.00/M output on Tinker; weights free under Apache 2.0; free route on OpenRouter.
Common questions
Inkling 2 Small FAQs
Is Inkling 2 Small really open source?
Yes. The full weights are published on Hugging Face under the Apache 2.0 license, so you can download, fine-tune, and use them commercially.
How much does it cost to use?
The weights are free. Hosted API access on Tinker costs $0.30 per million input tokens and $1.00 per million output tokens, and OpenRouter offers a free tier.
How big is the model, and can I run it at home?
It has 594B total parameters with 17B active per token. The NVFP4 checkpoint needs about 500GB of combined GPU memory, so most people use a hosted API.
What is it best at?
Agentic coding, knowledge work, and visual tasks. Artificial Analysis scores it 31 on its Intelligence Index, level with Gemini 3.8 Flash.
Where can I try it for free?
OpenRouter's free route (thinkingmachines/inkling-2-small:free) or the Tinker Playground, both with no payment needed to start.
Reviews
See what the community thinks and share your experience.
—
Based on 0 ratings
Rating distribution
Community reviews
No reviews
No written reviews yet.
