Reasoning models
Large AI models designed to generate intermediate reasoning steps before answering, aiming to solve problems like math proofs.
As defined in cwt_alison-gopnik
Other definitions
Models that spend more inference-time compute 'thinking' through an extended chain of thought before producing output, shifting cost from training to runtime.
A New Cold War, Rise of China and Global Chaos: How AI Will Rewrite Geopolitics | Gregory Smith
Where it comes up · 11
- Satya Nadella on AI Agents, Rebuilding the Web, the Future of Work, and more5 claims
- The AI Reset Is Here: Search, Jobs, and Everything Else w/ Anish Acharya, Dave Blundin, Salim Ismail4 claims
- Bret Taylor: Facebook, Salesforce, OpenAI, and Sierra4 claims
- A New Cold War, Rise of China and Global Chaos: How AI Will Rewrite Geopolitics | Gregory Smith3 claims
- Dylan Field - Designing The Future - [Invest Like the Best, EP.407]3 claims
- Gavin Baker - Watts and Wafers - [Invest Like the Best, EP.473]2 claims
- Yann LeCun Says LLMs Have 2 Years Left…1 claim
- Sam Altman talks the NYT lawsuit, Meta's talent poaching, and Trump on AI | Interview1 claim
and 3 more
Most important claims · 10 of 26Search all claims for “Reasoning models” →
- Reasoning models (trained via reinforcement learning on specific domains with formal correctness criteria) represent an innovation as significant as language models, but require domain-specific training datasets and model architectures for each domain before achieving broad capabilities like those predicted for GPT-5.The AI Reset Is Here: Search, Jobs, and Everything Else w/ Anish Acharya, Dave Blundin, Salim Ismail
- The Cold War II scenario rests on a technical assumption that AGI development remains very expensive, so that large firms and capital-rich states (with talent and educational capability) have an advantage; if AGI development turns out to be cheap—via leveling cost curves, reasoning models that run longer rather than train longer, or distilling cheaper models from a frontier model's outputs—then fast-followers could proliferate the technology and undermine the bipolar scenario.A New Cold War, Rise of China and Global Chaos: How AI Will Rewrite Geopolitics | Gregory Smith
- Reasoning models like OpenAI's O1 change the economics of AI by moving compute from training time to inference time, allowing model distillation to smaller models, which reduces the capital requirements for building and deploying frontier AI systems.Bret Taylor: Facebook, Salesforce, OpenAI, and Sierra
- Reasoning models (trained via reinforcement learning on specific domains) are as important an innovation as language models themselves, but require substantial domain-specific data collection before they can scale broadly.The AI Reset Is Here: Search, Jobs, and Everything Else w/ Anish Acharya, Dave Blundin, Salim Ismail
- The Cold War II scenario depends on AGI remaining very expensive to build, giving large firms and states the advantage; if development becomes cheap—through leveling compute cost curves, reasoning models that run longer rather than train longer, or distilling cheaper models from a leader's outputs—then AGI proliferates widely and the scenario breaks down.Rise And Fall Of Nations How Superintelligence Will Transform Geopolitics Gregory Smith
- DeepSeek was the first time it became clear how much more compute reasoning models consume during inference than non-reasoning models.Gavin Baker - Watts and Wafers - [Invest Like the Best, EP.473]
- Reasoning models (models that spend more computational time thinking through problems rather than producing immediate outputs) could reduce AGI development costs if they rely more on inference-time compute than training-time compute, enabling cheaper model advancementA New Cold War, Rise of China and Global Chaos: How AI Will Rewrite Geopolitics | Gregory Smith
- Reasoning models are an innovation in which a model 'thinks' more—engaging an extended chain of thought and spending more compute at inference time before producing an output—rather than simply producing a single output, which shifts cost from expensive training toward longer runtime.A New Cold War, Rise of China and Global Chaos: How AI Will Rewrite Geopolitics | Gregory Smith
- The primary inputs limiting AGI progress are data (textual training data is 'waning' with a potential 'data wall'), compute (capital-intensive, with hardware costs that may not be recovered), and algorithms (with potential plateaus in scaling laws even after breakthroughs like reasoning models).Bret Taylor: Facebook, Salesforce, OpenAI, and Sierra
- The sustainable competitive advantage for enterprises using AI is not static proprietary data, but rather the ability to create a virtuous feedback loop: take commoditized foundation models, fine-tune them with internal knowledge and data, deploy outputs, capture market signals (customer feedback, marketplace validation), and reinforce the model through reinforcement learning in the real world.Satya Nadella on AI Agents, Rebuilding the Web, the Future of Work, and more