Inkling 2 Small

Thinking Machines Lab released Inkling 2 Small on 5 November 2026, and for now the place to use it is Tinker, the company's own platform, with the weights held back. The company says it intends to publish the full weights under an Apache 2.0 licence "next week", once external safety testing and red-teaming are complete, and to reach other platforms in the coming weeks.

## A large model that keeps few parameters active

Inkling 2 Small is a mixture of experts with 594 billion parameters in total and 17 billion active. It takes image input as well as text, and it offers variable effort. The company states a 256K-token context window for its reinforcement learning run, and Tinker prices the model at both 64K and 256K contexts.

## Priced for fine-tuning first

The Tinker price list shows who the release is aimed at. It charges for prefill and sampling, and it has a separate training rate: $1.50 per million tokens at 64K context and $3.00 at 256K. Prefill costs $0.30, cached prefill $0.03 and sampling $1.00, at either context length. Serverless inference is listed as available in beta, though we did not read a price for it.

The company quotes an Artificial Analysis Intelligence Index score of 31 for the model. That figure comes from its own page.

## Key facts

**Maker:** Thinking Machines Lab **Released:** 5 November 2026 **Parameters:** 594B total, 17B active **Inputs:** text and images **Weights:** closed at release; Apache 2.0 promised after safety testing **Availability:** Tinker first, other platforms to follow

Not yet independently verified: Not yet independently verified. The specification and prices are the vendor's own and the weights are promised but not yet out; the vendor quotes an Artificial Analysis Intelligence Index score of 31, which we have not read on Artificial Analysis and do not store. We will update this when it can be confirmed, and remove this note.