Runtime as featured inForbesRead the article
All open source models
Thinking Machines Lab logo

Inkling 2 Small

Thinking Machines Lab · US · #3 in our ranking · Best US model

Inkling 2 Small is an open-weight mixture-of-experts model from Thinking Machines Lab (United States), released Nov 2026, with 594B parameters (17B active per token), a 1M-token context window, and the Apache 2.0.

Inkling 2 Small is a mixture-of-experts model that routes each token to 8 of 384 experts and mixes local and global attention. Announced on November 5, 2026, its weights went public on Hugging Face on November 12 in native NVFP4, with an MXFP8 alternative. It reads text and images and supports up to a 1M-token context.

Inkling 2 Small specs

Context1M tokens
Parameters594B
Provider
Thinking Machines Lab
Origin
US
Architecture
Mixture-of-experts
Parameters
594B total · 17B active
Context
1M tokens | 786K words
License
Apache 2.0 Commercial use
Input
Text, Image
Released
Nov 2026

Sourced from the model card on Hugging Face. Last checked 2026-11-21.

Highlights

  • 17B active parameters
  • Apache 2.0 license
  • Native NVFP4 weights, with an MXFP8 build

Where it fits for payment teams

A permissively licensed US model for tool-heavy agents, like investigations that call many APIs and MCP servers in one run.

Harnesses in Runtime

Run Inkling 2 Small through any of these harnesses on Runtime.

Claude CodeOpenCodeCline

Provider-reported benchmarks

DeepSWE v1.172.1
MCP Atlas85.4
BrowseComp90.2

Benchmark scores are reported by each model provider on its Hugging Face model card. They are not independently verified by Runtime.

01

Download the weights

Pull thinkingmachines/Inkling-2-Small from Hugging Face. Check the Apache 2.0.

02

Serve it in your cloud

vLLM on your GPUs, or a managed host.

03

Put it to work in Runtime

Point an agent or skill at it, then compare with evals.

Inkling 2 Small FAQ

What is the context window of Inkling 2 Small?

+

Inkling 2 Small supports a 1M-token context window (1,048,576 tokens), roughly 786K words of English text.

How many parameters does Inkling 2 Small have?

+

Inkling 2 Small is a mixture-of-experts model with 594B total parameters, of which 17B are active per token.

Is Inkling 2 Small free for commercial use?

+

Yes. Inkling 2 Small is released under the Apache 2.0 license, which is permissive and allows commercial use and fine-tuning.

What can Inkling 2 Small take as input?

+

Inkling 2 Small accepts text, image input and generates text.

Can I run Inkling 2 Small in my own cloud with Runtime?

+

Yes. Download the weights from Hugging Face (thinkingmachines/Inkling-2-Small) and serve them with vLLM in your cloud, or use a managed provider that hosts the model. Runtime agents can then use Inkling 2 Small through a harness like OpenCode, with your data staying in your environment. Our how-to-run page for Inkling 2 Small covers GPU memory, Bedrock, and Vertex AI.