These 129 models use chain-of-thought reasoning to break down complex problems step-by-step. They excel at math, logic, coding challenges, and multi-step planning — trading speed for accuracy on hard tasks.
Reasoning models "think out loud" before answering, breaking complex problems into smaller steps. This dramatically improves accuracy on math, logic puzzles, and multi-step coding tasks compared to standard models.
Reasoning takes extra tokens and time. These models are slower than standard models but significantly more accurate on hard problems. Use them when correctness matters more than latency.
Complex math and science problems, multi-step coding tasks, logical deduction, planning and strategy, and any task where standard models produce incorrect answers.
OpenAI's o1 and o3 series pioneered chain-of-thought reasoning. DeepSeek R1 demonstrated open-source reasoning. Many providers now offer reasoning-capable models at various price points.