Thinking Machines Lab releases Inkling 2 Mini, an open-weights looped-transformer model that runs on a laptop
Inkling 2 Mini at a glance Thinking Machines Lab released Inkling 2 Mini, an open-weights 4.3-billion-parameter looped-transformer model that runs on a laptop, costs $0.15 per million input tokens on Tinker and scores
Inkling 2 Mini at a glance
Thinking Machines Lab released Inkling 2 Mini, an open-weights 4.3-billion-parameter looped-transformer model that runs on a laptop, costs $0.15 per million input tokens on Tinker and scores 20.1 on Artificial Analysis's AI Index. The model will have 2.6 trillion training tokens at release, with a 1 million token context window and an Apache-2.0 license.
What changed with Inkling 2 Mini
Thinking Machines Lab announced on November 20, 2026, in its official announcement, that Inkling 2 Mini is a 4.3-billion-parameter model with a looped-transformer architecture: its 14 attention layers run repeatedly, alternating between attention and memory updates, for 22 cycles. The looped design means a small model can carry more capacity, so the model can run on a laptop and still reason well enough for agents, coding and interactive applications. The model is Apache-2.0 licensed, and its weights, model card and inference code are public on Hugging Face. It supports a context of up to 1 million tokens, but only text inputs, and it does not do image or audio inputs.
The model is available as a hosted service on Tinker, Thinking Machines Lab's inference platform. Tinker pricing lists Inkling 2 Mini at $0.15 per million input tokens, $0.50 per million output tokens and $0.75 per million cached input tokens.
Inkling 2 Mini benchmarks and pricing
Values are copied from the linked sources on the date shown. They are not IA Compa measurements.
- AI Index (Artificial Analysis) β 20.1 (as of Nov 20, 2026)
- Output speed (Artificial Analysis) β 279.9 tokens per second (as of Nov 20, 2026)
- Output tokens per second per dollar (Artificial Analysis) β 1,866.1 tokens per second per dollar (as of Nov 20, 2026)
- Input price per 1M tokens (Tinker) β $0.15 (as of Nov 20, 2026)
- Output price per 1M tokens (Tinker) β $0.50 (as of Nov 20, 2026)
- Cached input price per 1M tokens (Tinker) β $0.75 (as of Nov 20, 2026)
Where to compare Inkling 2 Mini
Leaderboards:
Official resources:
How it compares
On Artificial Analysis's AI Index, Inkling 2 Mini ranks #42 of 51 with 20.1, behind leader Claude Fable 5.2 at 64.6. It is the fastest model listed for output speed at 279.9 tokens per second, and the cheapest listed at $0.15 per million input tokens.
Why Inkling 2 Mini matters for businesses
Why it matters: a small business can run a real model inside its own product without sending customer data to a hosted API, at $0.15 per million input tokens on Tinker and on hardware it already owns. The 1 million token context means the model can hold a whole codebase or a long document in one prompt, so local agents and coding assistants can work on large projects without splitting them up.
Sources
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.