← Back to the Model Guide
Cheapest model / capability
Reasoning: the cheapest models
Models built for hard reasoning and planning.
How this page ranks
- Ranked by blended price per million tokens: three parts input to one part output, at the OpenRouter list price for the model. Lowest first. Ties go to the lower input price.
- Reasoning is a capability, not a job: these are models tagged reasoning on their record.
- Each model must list text output on OpenRouter or on its own record.
- Prices and scores come from the market tables only, each with the day it is for. The price on a model's own record is not used.
- Missing data never ranks. A model with no confirmed price in the last 30 days, a price in another unit, a $0 listing or unknown inputs and outputs is listed apart, with the reason.
Market data as of 10 Nov 2026. Prices: OpenRouter models API (official).
Ranked: 37 of 180 models
Best for reasoning →| # | Model | Input / 1M | Output / 1M | Blended / 1M | Price date |
|---|---|---|---|---|---|
| 1 |  DeepSeek V3.2DeepSeek | $0.260 | $0.380 | $0.290 | 10 Nov 2026 |
| 2 | ATrinity Large ThinkingArcee Ai | $0.250 | $0.800 | $0.388 | 23 Oct 2026 |
| 3 |  Qwen3 Next 80B A3B ThinkingQwen | $0.150 | $1.20 | $0.412 | 26 Oct 2026 |
| 4 | Qwen3 VL 30B A3B ThinkingQwen | $0.290 | $1.00 | $0.467 | 10 Nov 2026 |
| 5 | R1 0528DeepSeek | $0.500 | $2.15 | $0.912 | 10 Nov 2026 |
| 6 |  o3 MiniOpenAI | $0.550 | $2.20 | $0.963 | 10 Nov 2026 |
| 7 | o4 MiniOpenAI | $0.550 | $2.20 | $0.963 | 10 Nov 2026 |
| 8 |  Kimi K2 ThinkingMoonshotai | $0.600 | $2.50 | $1.07 | 10 Nov 2026 |
| 9 | Qwen3 235B A22B Thinking 2507Qwen | $0.450 | $3.50 | $1.21 | 10 Nov 2026 |
| 10 | Qwen3 235B A22B Thinking 2507 TEEQwen | $0.450 | $3.50 | $1.21 | 10 Nov 2026 |
| 11 |  Glm 5.2Cf | $0.0070 | $6.80 | $1.71 | 10 Nov 2026 |
| 12 | GLM 5.2Accounts | $0.0070 | $6.80 | $1.71 | 10 Nov 2026 |
| 13 | GLM-5.2Hf Zai Org | $0.0070 | $6.80 | $1.71 | 10 Nov 2026 |
| 14 | GLM-5.2Zai Org | $0.0070 | $6.80 | $1.71 | 10 Nov 2026 |
| 15 | o3 Mini HighOpenAI | $1.10 | $4.40 | $1.93 | 26 Oct 2026 |
| 16 | o4 Mini HighOpenAI | $1.10 | $4.40 | $1.93 | 26 Oct 2026 |
| 17 | GPT-6.1 TerraOpenAI | $1.00 | $5.00 | $2.00 | 10 Nov 2026 |
| 18 |  Grok 4.5xAI | $2.00 | $6.00 | $3.00 | 10 Nov 2026 |
| 19 |  Mistral LargeMistral | $2.00 | $6.00 | $3.00 | 10 Nov 2026 |
| 20 |  Gemini 2.5 ProGoogle | $1.25 | $10.00 | $3.44 | 10 Nov 2026 |
| 21 | o3OpenAI | $2.00 | $8.00 | $3.50 | 10 Nov 2026 |
| 22 |  Sonar Reasoning ProPerplexity | $2.00 | $8.00 | $3.50 | 10 Nov 2026 |
| 23 |  Claude Sonnet 5.5Anthropic | $2.00 | $10.00 | $4.00 | 10 Nov 2026 |
| 24 | Gemini 4 ArgonGoogle | $2.00 | $10.00 | $4.00 | 10 Nov 2026 |
| 25 | GPT-6.1 SolOpenAI | $2.00 | $10.00 | $4.00 | 10 Nov 2026 |
Showing the top 25 of 37 ranked models.
Not ranked, and why (143)
No confirmed price in the market tables in the last 30 days 143
- Arcee AI: Maestro Reasoning Not matched to an OpenRouter model
- Claude 4.6 Opus Thinking Not matched to an OpenRouter model
- Claude 4.6 Opus Thinking Low Not matched to an OpenRouter model
- Claude 4.6 Opus Thinking Max Not matched to an OpenRouter model
- Claude 4.6 Opus Thinking Medium Not matched to an OpenRouter model
- Claude 4.7 Opus Thinking Not matched to an OpenRouter model
- Claude Fable 5 (batch) Not matched to an OpenRouter model
- Claude Fable 5.1 (batch) Not matched to an OpenRouter model
- Claude Fable 5.2 (batch) Not matched to an OpenRouter model
- Claude Opus 4 Not matched to an OpenRouter model
- Claude Opus 4.7 (Fast) Not matched to an OpenRouter model
- Claude Opus 4.8 (Fast) Not matched to an OpenRouter model
- Claude Opus 4.8 Thinking Not matched to an OpenRouter model
- Claude Opus 5.5 (batch) Not matched to an OpenRouter model
- Claude Sonnet 4 Not matched to an OpenRouter model
- Claude Sonnet 4.6 Thinking Not matched to an OpenRouter model
- Claude Sonnet 5 (Free) Not matched to an OpenRouter model
- Claude Sonnet 5.5 (batch) Not matched to an OpenRouter model
- DeepSeek R1 Not matched to an OpenRouter model
- DeepSeek R1 (Turbo) Not matched to an OpenRouter model
- DeepSeek R1 0528 Qwen3 8B Not matched to an OpenRouter model
- DeepSeek R1 Distill Qwen 14B Not matched to an OpenRouter model
- Deepseek R1 Distill Qwen 32B Not matched to an OpenRouter model
- DeepSeek R1 Distill Qwen 32B Not matched to an OpenRouter model
- DeepSeek R1 Llama 70B Abliterated Not matched to an OpenRouter model
- DeepSeek R1 Qwen Abliterated Not matched to an OpenRouter model
- DeepSeek Reasoner Not matched to an OpenRouter model
- DeepSeek V3.1 Terminus (Thinking) Not matched to an OpenRouter model
- DeepSeek V3.1 Thinking Not matched to an OpenRouter model
- DeepSeek V3.2 Thinking Not matched to an OpenRouter model
- DeepSeek V4 Flash (Thinking) Not matched to an OpenRouter model
- DeepSeek V4 Pro (Thinking) Not matched to an OpenRouter model
- DeepSeek V4 Pro Cheaper (Thinking) Not matched to an OpenRouter model
- DeepSeek V4 Pro Thinking TEE Not matched to an OpenRouter model
- DeepSeek/DeepSeek-V3.2-Exp-Thinking Not matched to an OpenRouter model
- ERNIE 5.0 Not matched to an OpenRouter model
- ERNIE-4.5-21B-A3B-Thinking Not matched to an OpenRouter model
- ERNIE-4.5-VL-28B-A3B-Thinking Not matched to an OpenRouter model
- Gemini 3 Flash Thinking Not matched to an OpenRouter model
- Gemini 3.5 Flash Thinking Not matched to an OpenRouter model
- Gemini 4 Argon (batch) Not matched to an OpenRouter model
- Gemma 4 26B A4B Thinking Not matched to an OpenRouter model
- Gemma 4 31B Thinking Not matched to an OpenRouter model
- Gemma 4 31B Thinking TEE Not matched to an OpenRouter model
- GLM 4.5 (Thinking) Not matched to an OpenRouter model
- GLM 4.5 Air (Thinking) Not matched to an OpenRouter model
- GLM 4.5V Thinking Not matched to an OpenRouter model
- GLM 4.6 Thinking Not matched to an OpenRouter model
- GLM 4.6 Turbo (Thinking) Not matched to an OpenRouter model
- GLM 4.7 Flash Original Thinking Not matched to an OpenRouter model
- GLM 4.7 Flash Thinking Not matched to an OpenRouter model
- GLM 4.7 Original Thinking Not matched to an OpenRouter model
- GLM 4.7 Thinking Not matched to an OpenRouter model
- GLM 5 Original Thinking Not matched to an OpenRouter model
- GLM 5 Thinking Not matched to an OpenRouter model
- GLM 5.1 Thinking Not matched to an OpenRouter model
- GLM 5.1 Thinking TEE Not matched to an OpenRouter model
- GLM 5V Turbo Thinking Not matched to an OpenRouter model
- GLM-5.2 Not matched to an OpenRouter model
- GPT 5.1 Thinking Not matched to an OpenRouter model
- GPT-6 Astra (batch) Not matched to an OpenRouter model
- GPT-6.1 Sol (batch) Not matched to an OpenRouter model
- GPT-6.1 Terra (batch) Not matched to an OpenRouter model
- Grok 4.1 Fast Non-Reasoning Not matched to an OpenRouter model
- Grok 4.1 Fast Reasoning Not matched to an OpenRouter model
- Grok 4.2 Fast Non Reasoning Not matched to an OpenRouter model
- Grok 4.20 (Reasoning) Not matched to an OpenRouter model
- Grok 4.20 Beta Non-Reasoning Not matched to an OpenRouter model
- Grok 4.20 Beta Reasoning Not matched to an OpenRouter model
- Grok 4.20 Non-Reasoning Not matched to an OpenRouter model
- Grok 4.20 Reasoning Not matched to an OpenRouter model
- Grok-4-Fast-Non-Reasoning Not matched to an OpenRouter model
- Grok-4-Fast-Reasoning Not matched to an OpenRouter model
- Hermes 4 (Thinking) Not matched to an OpenRouter model
- Hermes 4 Large (Thinking) Not matched to an OpenRouter model
- Kimi K2 Thinking Original Not matched to an OpenRouter model
- Kimi K2 Thinking Turbo Not matched to an OpenRouter model
- Kimi K2 Thinking Turbo Original Not matched to an OpenRouter model
- Kimi K2.5 Thinking Not matched to an OpenRouter model
- Kimi K2.5 Thinking TEE Not matched to an OpenRouter model
- Kimi K2.6 Thinking Not matched to an OpenRouter model
- kimi-k2-thinking Not matched to an OpenRouter model
- LFM2.5-1.2B-Thinking (free) Not matched to an OpenRouter model
- Llama 3.3 70B Cu Mai Not matched to an OpenRouter model
- LongCat Flash Thinking 2601 Not matched to an OpenRouter model
- Mag Mell R1 Not matched to an OpenRouter model
- Magistral Medium (latest) Not matched to an OpenRouter model
- Magistral Medium (latest) Not matched to an OpenRouter model
- Magistral Small Not matched to an OpenRouter model
- Magistral Small 2506 Not matched to an OpenRouter model
- MAI-DS-R1 Not matched to an OpenRouter model
- MiMo V2 Flash (Thinking) Not matched to an OpenRouter model
- MiMo V2 Flash (Thinking) Original Not matched to an OpenRouter model
- MiniMax M3 Thinking Not matched to an OpenRouter model
- Ministral 3 14B Reasoning 2512 Not matched to an OpenRouter model
- Ministral 3 3B Reasoning 2512 Not matched to an OpenRouter model
- Mistral Medium 3.5 Thinking Not matched to an OpenRouter model
- Mistral Small 4 119B Thinking Not matched to an OpenRouter model
- Nemotron 3 Nano Omni (free) Not matched to an OpenRouter model
- Nemotron 3 Nano Omni 30B A3B Reasoning Not matched to an OpenRouter model
- nemotron-content-safety-reasoning-4b Not matched to an OpenRouter model
- NVIDIA Nemotron 3 Nano Omni Not matched to an OpenRouter model
- Nvidia Nemotron 3 Super 120B Thinking Not matched to an OpenRouter model
- o3 (batch) Not matched to an OpenRouter model
- o3 Deep Research Not matched to an OpenRouter model
- o3 Mini (batch) Not matched to an OpenRouter model
- o4 Mini (batch) Not matched to an OpenRouter model
- o4 Mini Deep Research Not matched to an OpenRouter model
- OpenAI o1-mini Not matched to an OpenRouter model
- OpenAI o1-preview Not matched to an OpenRouter model
- OpenAI o3-mini (Low) Not matched to an OpenRouter model
- OpenAI o3-pro (2025-06-10) Not matched to an OpenRouter model
- Phi-4-mini-reasoning Not matched to an OpenRouter model
- Phi-4-Reasoning Not matched to an OpenRouter model
- Pro/deepseek-ai/DeepSeek-R1 No confirmed price for deepseek/deepseek-r1
- Qwen Plus 0728 (thinking) Not matched to an OpenRouter model
- Qwen QwQ 32B Preview Not matched to an OpenRouter model
- Qwen/Qwen3-VL-32B-Thinking Not matched to an OpenRouter model
- Qwen3 235B A22B Thinking 2507 Not matched to an OpenRouter model
- Qwen3 30B A3B Thinking 2507 Not matched to an OpenRouter model
And 23 more.
Automated from dated market data, not Enterprise DNA tested picks. Check the price and terms with the provider before you commit. Methodology: how the model numbers work.