Reasoning models
296 models that reason step by step before they answer, for math, code and multi-step problems.
/
 Gemini 4 ArgonGooglePrompt cachingJSON+3 | $8/1Mvia APIMart | 1.05M | $2/1M | $10/1M | 1673 | 2M | |
| Kimi K3Moonshot AIOpen weights · with termsPrompt caching+4 | $8.28/1Mvia Kenari | 1.05M | $2.78/1M | $13.89/1M | 1655 | 658K | |
| GPT-6 AstraOpenAIPrompt cachingJSON+3 | $25/1Mvia CrazyRouter | 1.05M | $10/1M | $50/1M | 1775 | 580K | |
 Muse WonderMetaPrompt cachingJSON+3 | $16/1Mvia CometAPI | 1.05M | $3.20/1M | $16/1M | — | 569K | |
| Claude Opus 5.5AnthropicPrompt cachingJSON+3 | $16/1Mvia APIMart | 1M | $4/1M | $20/1M | 1807 | 405K | |
| GPT-6.1 SolOpenAIPrompt cachingJSON+3 | $8/1Mvia APIMart | 1.05M | $2/1M | $10/1M | 1749 | 384K | |
 GLM-5.3Zhipu AIOpen weights · with termsPrompt caching+3 | $1.10/1Mvia NanoGPT | 1.05M | $1.40/1M | $4.40/1M | 1623 | 276K | |
| GPT-6 SolOpenAIPrompt cachingJSON+3 | $8/1Mvia APIMart | 1.05M | $2/1M | $10/1M | 1682 | 242K | |
 Qwen3.8 2.4T A95BAlibabaOpen weights · with termsPrompt caching+3 | $4.11/1Mvia APIMart | 1M | $2/1M | $6/1M | — | 234K | |
| Gemini 3 FlashGooglePrompt cachingJSON+3 | $2.40/1Mvia CometAPI | 1.05M | $0.30/1M | $2.40/1M | 1466 | 207K | |
 DeepSeek V4.1 FlashDeepSeekMITPrompt caching+4 | $0.03/1Mvia UnoRouter | 1.05M | $0.007/1M | $0.03/1M | 1664 | 138K | |
 Mistral Large 4MistralPrompt cachingJSON+3 | $2.09/1Mvia Opper | 1.05M | $1.36/1M | $4.18/1M | 1543 | 122K | |
| GLM-5.3-FlashZhipu AIMITPrompt caching+4 | $0.20/1Mvia APIMart | 1.05M | $0.15/1M | $0.50/1M | 1608 | 121K | |
| Muse Spark 1.3MetaPrompt cachingJSON+3 | $4.25/1Mvia EmpirioLabs AI | 1.05M | $1.25/1M | $4.25/1M | 1657 | 121K | |
| Qwen3.8 27BAlibabaApache 2.0Prompt caching+4 | $0.50/1Mvia EmpirioLabs AI | 262K | $0.50/1M | $3/1M | 1591 | 120K | |
| GPT-6 LunaOpenAIPrompt cachingJSON+3 | $0.40/1Mvia APIMart | 1.05M | $0.10/1M | $0.50/1M | 1578 | 117K | |
| GLM-5.2Zhipu AIMITPrompt caching+3 | $1.80/1Mvia Deep Infra | 1M | $1.40/1M | $4.40/1M | 1604 | 109K | |
| Claude Sonnet 5.5AnthropicPrompt cachingJSON+3 | $8/1Mvia APIMart | 1M | $2/1M | $10/1M | 1767 | 102K | |
| Gemini 3.8 FlashGooglePrompt cachingJSON+3 | $3/1Mvia APIMart | 1.05M | $0.75/1M | $3.75/1M | 1585 | 102K | |
| Claude Haiku 5.5AnthropicPrompt cachingJSON+3 | $0.30/1Mvia Kie.ai | 1M | $0.10/1M | $0.50/1M | 1585 | 94K | |
| Muse Spark OpenMetaPrompt cachingJSON+3 | $4.25/1Mvia Baseten | 1.05M | $1.25/1M | $4.25/1M | 1527 | 74K | |
 MiMo-V2.6-ProXiaomiPrompt cachingJSON+3 | $0.696/1Mvia CometAPI | 1.05M | $0.348/1M | $0.696/1M | 1625 | 72K | |
| Qwen3.8 Flash NextAlibabaOpen weights · with termsPrompt caching+4 | $0.376/1Mvia CometAPI | 262K | $0.12/1M | $0.376/1M | 1628 | 63K | |
| DeepSeek V4 FlashDeepSeekMITPrompt caching+3 | $0.17/1Mvia Opper | 1.05M | $0.07/1M | $0.17/1M | — | 51K | |
| DeepSeek V4 Flash Vision ExpDeepSeekMITPrompt caching+4 | $0.294/1Mvia Deep Infra | 1.05M | $0.15/1M | $0.60/1M | — | 47K | |
| Claude Opus 5AnthropicPrompt cachingJSON+3 | $20/1Mvia APIMart | 1M | $5/1M | $25/1M | 1690 | 46K | |
| DeepSeek V4 ProDeepSeekMITPrompt caching+3 | $0.418/1Mvia OpenRouter | 1.05M | $0.209/1M | $0.418/1M | — | 40K | |
 Ling 3.1 FlashInclusionaiMITPrompt caching+2 | $0.225/1Mvia Deep Infra | 262K | $0.075/1M | $0.225/1M | — | 38K | |
| Claude Sonnet 4.6AnthropicPrompt cachingJSON+3 | $12/1Mvia APIMart | 1M | $3/1M | $15/1M | 1523 | 36K | |
| GPT-5OpenAIPrompt cachingJSON+3 | $8/1Mvia APIMart | 400K | $1.25/1M | $10/1M | 1418 | 35K | |
| Claude Fable 5AnthropicPrompt cachingJSON+3 | $40/1Mvia APIMart | 1M | $10/1M | $50/1M | 1624 | 35K | |
| GPT OSS 120BOpenAIApache 2.0Prompt caching+3 | $0.14/1Mvia Requesty | 131K | $0.032/1M | $0.14/1M | 1366 | 35K | |
| Claude Fable 5.1AnthropicPrompt cachingJSON+3 | $40/1Mvia APIMart | 1M | $10/1M | $50/1M | 1743 | 34K | |
| GPT-5.6 LunaOpenAIPrompt cachingJSON+3 | $0.96/1Mvia APIMart | 1.05M | $0.20/1M | $1.20/1M | 1519 | 33K | |
 MiniMax-M3.1-FlashMiniMaxOpen weights · with termsPrompt caching+4 | $0.45/1Mvia EmpirioLabs AI | 1.05M | $0.15/1M | $0.60/1M | 1590 | 29K | |
| MiMo-V2.6-FlashXiaomiPrompt cachingJSON+3 | $0.10/1Mvia UnoRouter | 1.05M | $0.05/1M | $0.10/1M | 1634 | 28K | |
| GPT-5.6OpenAIPrompt cachingJSON+3 | $10/1Mvia Boundless | 1.05M | $4/1M | $20/1M | 1615 | 28K | |
| MiniMax-M3MiniMaxOpen weights · with termsPrompt caching+4 | $0.90/1Mvia EmpirioLabs AI | 1M | $0.30/1M | $1.20/1M | 1482 | 26K | |
| Qwen3.8 MaxAlibabaPrompt cachingJSON+3 | $4.11/1Mvia APIMart | 1M | $1.65/1M | $4.95/1M | 1673 | 25K | |
| GLM-5Zhipu AIMITPrompt caching+3 | $1.28/1Mvia APIMart | 205K | $1/1M | $3.20/1M | 1446 | 23K | |
| Grok 3xAIPrompt cachingJSON+2 | $15/1Mvia AIHubMix | 131K | $3/1M | $15/1M | — | 23K | |
| DeepSeek R1DeepSeekMITReasoning+1 | $1.79/1Mvia APIMart | 164K | $0.448/1M | $1.79/1M | — | 22K | |
| Kimi K2.5Moonshot AIOpen weights · with termsPrompt caching+4 | $0.555/1Mvia APIMart | 262K | $0.411/1M | $0.555/1M | 1445 | 22K | |
| Claude Sonnet 5AnthropicPrompt cachingJSON+3 | $6.90/1Mvia Kenari | 1M | $2/1M | $10/1M | 1541 | 22K | |
| Qwen3.8 FlashAlibabaPrompt cachingJSON+3 | $0.309/1Mvia APIMart | 1M | $0.113/1M | $0.382/1M | — | 20K | |
| Qwen3.5 9BAlibabaApache 2.0Prompt caching+4 | $0.13/1Mvia EmpirioLabs AI | 262K | $0.09/1M | $0.13/1M | — | 19K | |
| LongCat 2.5 PreviewMeituanPrompt cachingJSON+3 | $1.18/1Mvia ZenMux | 1.05M | $0.30/1M | $1.18/1M | — | 19K | |
| Grok 4.20xAIPrompt cachingJSON+3 | $2.50/1Mvia APIYI | 1M | $1.25/1M | $2.50/1M | — | 15K | |
| Grok-4xAIPrompt cachingJSON+3 | $14.25/1Mvia Jiekou.AI | 256K | $2.85/1M | $14.25/1M | 1411 | 15K | |
| Gemini 2.5 ProGooglePrompt cachingJSON+3 | $8/1Mvia APIMart | 1.05M | $1.25/1M | $10/1M | 1458 | 14K |
Frequently asked questions
A model trained to work through a problem in intermediate steps, often called thinking, before it answers. It does better on math, coding and planning, at the cost of more tokens and slower replies.
The thinking steps are billed as output tokens, so one answer can use many times the tokens of a direct reply. Compare output prices, and check whether the model lets you limit or switch off its reasoning.
For short lookups, classification, extraction or chat, a regular model is faster and cheaper with little loss in quality. Keep reasoning for problems that really need several steps.
Can't find what you're looking for? Contact us