Reasoning models
AI models that "think" longer to solve harder problems step by step.
Ranks moderate (58/100) — set apart by rising momentum (78) and recent news intensity (26).
Why it matters
They change the balance between training and inference demand.
Why now
Reasoning models trade more inference compute for capability, reshaping where compute is consumed.
If Reasoning models runs short
Runaway inference cost makes reasoning uneconomic for some uses.
In depth · editorial + model
Reasoning models are AI systems that "think" for longer — working through a problem step by step at answer time rather than responding in a single pass. They matter because they shift the balance between training and inference: instead of baking all capability in during pretraining, they buy additional capability by spending more compute at the moment of inference. That reshapes where compute is consumed in the stack, pulling demand toward serving rather than training.
The catch is cost: because reasoning burns far more inference compute per answer, runaway serving costs can make it uneconomic for some uses, and the technique only pays off where the harder answer is worth the extra spend. Capability, in this model, can be bought at inference time — if you can afford it. The exposed names are the labs pushing reasoning hardest — DeepSeek, Zhipu AI, Moonshot AI, xAI, Mistral AI, and Safe Superintelligence — and compute-heavy backers such as High-Flyer Quant.
Who makes Reasoning models
The companies exposed to Reasoning models
How to think about it
- Capability can be bought at inference time
What to watch
- Inference compute per query
- Reasoning cost/accuracy trade-offs
Frequently asked
What is Reasoning models?
AI models that "think" longer to solve harder problems step by step.
Why does Reasoning models matter for AI?
They change the balance between training and inference demand.
Who makes Reasoning models?
The companies the model tags as producers or suppliers of Reasoning models: DeepSeek, Zhipu AI (Z.ai), Moonshot AI, xAI, Mistral AI.
Which companies are most exposed to Reasoning models?
DeepSeek, Zhipu AI (Z.ai), Moonshot AI, Safe Superintelligence, xAI, High-Flyer Quant — 7 companies in total are mapped to Reasoning models.
What happens if Reasoning models runs short?
Runaway inference cost makes reasoning uneconomic for some uses.
Where does Reasoning models sit in the AI value chain?
Reasoning models sits in the Models layer of the AI value chain.
Go deeper on Reasoning models
- The materials, geographies and policies it depends on — heat-mapped
- Substitutes, relief valves and the domino chains if it tightens
- The live tension score, momentum and news drivers
- Four levels of analysis — from plain-English to strategic
model v0.7.0 · research, not advice
Model scores are illustrative reads from our model of the AI value chain — not investment advice.