Reasoning AI Models — Pricing & Comparison
Specialized models that use chain-of-thought to solve complex math, logic, coding, and analytical problems step by step.
All Reasoning Models — Sorted by Price
| # | Model | Provider | Input $/M | Output $/M | Context | 1M Requests* |
|---|---|---|---|---|---|---|
| 1 | DeepSeek R1 V3.2 2025-01-20 | DeepSeek | $0.28 | $0.42 | 128K | $$0.000490 |
| 2 | Magistral Small 2025-12-15 | Mistral AI | $0.5 | $1.5 | 128K | $$0.001250 |
| 3 | o4-mini 2025-04-16 | OpenAI | $1.1 | $4.4 | 2000K | $$0.003300 |
| 4 | o3-mini 2025-01-31 | OpenAI | $1.1 | $4.4 | 500K | $$0.003300 |
| 5 | o1-mini 2024-09-12 | OpenAI | $1.1 | $4.4 | 128K | $$0.003300 |
| 6 | o4-mini Deep Research 2025-06-26 | OpenAI | $2 | $8 | 200K | $$0.006000 |
| 7 | o3 2025-04-16 | OpenAI | $2 | $8 | 1000K | $$0.006000 |
| 8 | Magistral Medium 2025-12-15 | Mistral AI | $2 | $5 | 128K | $$0.004500 |
| 9 | Grok 4.20 2026-02-17 | xAI | $2 | $6 | 2000K | $$0.005000 |
| 10 | Grok 4 2025-07-09 | xAI | $3 | $15 | 256K | $$0.0105 |
| 11 | o3 Deep Research 2025-06-26 | OpenAI | $10 | $40 | 200K | $$0.03 |
| 12 | o1 2024-09-12 | OpenAI | $15 | $60 | 200K | $$0.045 |
| 13 | o3-pro 2025-06-10 | OpenAI | $20 | $80 | 1000K | $$0.06 |
| 14 | GPT-5.2 pro 2025-12-11 | OpenAI | $21 | $168 | 1000K | $$0.105 |
| 15 | GPT-5.4 Pro 2026-03-06 | OpenAI | $30 | $180 | 1050K | $$0.12 |
* Estimated cost for 1M requests at 1,000 input + 500 output tokens each.
Calculate Your Reasoning Model Costs
💡 Tips for Using Reasoning Models
Reasoning models excel at multi-step problems — math proofs, code debugging, logical analysis
They often use more output tokens due to chain-of-thought, so factor that into cost estimates
For simple questions, a balanced model may give equally good answers at lower cost
Compare Reasoning Models
Frequently Asked Questions
What are reasoning models?
Reasoning models are trained to "think out loud" using chain-of-thought. They break complex problems into steps, making them better at math, coding, logic puzzles, and multi-step analysis.
Why do reasoning models cost more per request?
Reasoning models generate more tokens (their thinking steps) before producing the final answer. This means higher output token usage per request, even if the final answer is short.
When should I use a reasoning model vs a flagship model?
Use reasoning models for problems with clear right/wrong answers (math, code, logic). Use flagship models for creative, open-ended, or nuanced tasks.