Thinking Machines LabLanguage modelActive
Inkling 2 Small
Summary
Inkling 2 Small is the first model of Thinking Machines Lab's second generation, released on November 5, 2026. It is a mixture-of-experts model with 594 billion total parameters, 17 billion of them active. It accepts text and images, answers in text, and offers adjustable effort. According to Thinking Machines it scores 31 on the Artificial Analysis Intelligence Index, well above the lab's earlier models.
Access starts on Tinker, for inference and fine-tuning, with closed weights under the Tinker Terms of Use. Thinking Machines plans to release the full weights under Apache 2.0 after external safety testing; that license does not yet cover the current weights. The model card cites up to 1M tokens of context, while Tinker's standard variant lists 64K and a separate variant 256K. Sampling on Tinker costs $1.00 per million tokens.
All specifications
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
- Access
- Available on Tinker for inference and fine-tuning. Tinker Playground provides inference access.
- Source: Access
Further details
- Architecture and size
- Mixture-of-experts model with 594 billion total parameters and 17 billion active parameters.
- Source: Architecture and size
- License and weights
- The current model card lists the Tinker Terms of Use. The weights are currently closed.
- Source: License and weights
- Announced weights plan
- Thinking Machines intends to release the full weights under Apache 2.0 after external safety testing and red-teaming. This is not a license for the currently closed weights.
- Source: Announced weights plan
- Model-card context statement
- The model card states a context window of up to 1M tokens.
- Source: Model-card context statement
- Tinker standard variant
- Tinker's pricing list shows 64K context for thinkingmachines/Inkling-2-Small.
- Source: Tinker standard variant
- Separate Tinker variant
- thinkingmachines/Inkling-2-Small:peft:262144 is the separate 256K variant. The 262144 suffix belongs to this variant and does not establish the token count of the standard endpoint.
- Source: Separate Tinker variant
Prices
- Tinker standard variant, prefill
- Tinker prefill price: $0.30 per 1M tokens, or $0.03 for cached input.Source
- Tinker standard variant, sampling and training
- Tinker sampling price: $1.00 per 1M tokens. Tinker training price: $1.50 per 1M tokens.Source
- 256K Tinker variant, prefill
- For thinkingmachines/Inkling-2-Small:peft:262144, Tinker also lists $0.30 prefill and $0.03 cached input per 1M tokens.Source
- 256K Tinker variant, sampling and training
- For thinkingmachines/Inkling-2-Small:peft:262144, Tinker lists $1.00 sampling and $3.00 training per 1M tokens.Source
Frequently asked questions about Inkling 2 Small
Can I run Inkling 2 Small locally?
No. Thinking Machines Lab does not publish the weights.
Stated by Thinking Machines Lab
- Open weights
- no
- License
- n/a
- Parameters
- 594B
- active per token
- 17B