New to this? Read it in simple words
- Thinking Machines released Inkling 2 Small on November 5, on its Tinker service.
- An outside test group gave it 31, level with Gemini 3.8 Flash (high).
- The weights are not public yet. The company plans them for next week.
- It costs more per task than before because it writes much more.
- Weights
- The numbers a model learned in training. With them, people can run the model themselves.
- Mixture-of-Experts
- A model design in which only part of the model works on each step.
- Tokens
- Small pieces of text that AI models read and write, and that services charge for.
Inkling 2 Small is the second generation of Thinking Machines’ own models. The company’s first product, Tinker, helps people fine-tune AI models, which means adapting a model for one kind of work.
What it is
The company calls it a Mixture-of-Experts model with 594 billion total parameters, of which 17 billion are active. Only part of such a model works on each step, which keeps it cheaper to run. It takes text and images as input.
It costs $0.30 per million input tokens and $1.00 per million output tokens, Artificial Analysis reports. People can use it on Tinker and in Tinker Playground. The company says other platforms will follow in the coming weeks.
Scores and task costs are from Artificial Analysis. The weights plan is the company’s own.
The outside test
Artificial Analysis gave the model a score of 31 on its Intelligence Index, 18 points above Inkling Small. The test group says that matches Gemini 3.8 Flash (high) and beats Xiaomi’s MiMo-V2.6-Pro.
The results against Claude Sonnet 5.5 (medium) are mixed. Inkling 2 Small scored 25% to 11% on FrontierCode 2.0, but 13% to 30% on Terminal-Bench 4.0.
Its rate of made-up answers fell from 61% to 27% on one test. That is mostly because it answers less often: its attempt rate fell from 77% to 53%, and its accuracy rose only from 32% to 34%.
The real cost of a job
The model writes about 150,000 output tokens per test task, about four times as many as Inkling Small. So its cost per task rises from $0.20 to $1.18, Artificial Analysis reports.
Two rivals still cost far more on the same test: $7.77 for Gemini 3.8 Flash (high) and $10.02 for Claude Sonnet 5.5 (medium).
Weights, safety and money
“We intend to make the full weights available under an Apache 2.0 license next week after the completion of external safety testing and red-teaming,” the company wrote. Its own Inkling page says “in the coming weeks”, and Artificial Analysis calls the date unconfirmed.
Today the model card lists the Tinker Terms of Use as its licence. The card says outside testers were involved, and that loss-of-control tests found the model “materially below frontier capabilities”. It names no testers, so these are the company’s own statements.
Separately, Business Insider reports that the company is raising $5 billion at a $40 billion valuation before the new money, led by Accel. That rests on unnamed sources. Thinking Machines declined to comment, and Accel did not respond.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1
- 2
- 3Research · November 5, 2026Thinking Machines Inkling 2 Small: the highest scoring open weights model outside of China pending weights release Artificial Analysis
- 4Research · November 6, 2026Thinking Machines’ Inkling 2 Small claims top spot among western open-weight models The Decoder
- 5Research · November 6, 2026Thinking Machines Launches Inkling 2 Small as Reported $5B Funding Round Nears Techstrong.ai
- 6Research · November 5, 2026Thinking Machines Lab Is Closing in on a $5 Billion New Funding Round Business Insider
The scores and task costs come from Artificial Analysis, an outside test group. The size, the safety work and the plan for the weights are the company’s own claims. Only Business Insider reports the funding round.
Thinking Machines launches Inkling 2 Small, weights to come. Silicon AI News, November 8, 2026. https://siliconainews.com/stories/thinking-machines-inkling-2-small
A strong new model that matches Gemini 3.8 Flash at far lower task cost, though its open weights are still only a promise. How we rate the mood
The mood shows how the news frames the event for the tech and AI world: as an opportunity (6 and above) or as a loss (below 5). It is not our verdict on who is right.