Open-source vs closed model: total cost compared
Teams weighing self-hosting often need to compare total cost of an open self-hosted model versus a closed API. Because providers price input and output tokens very differently, the "obvious" model is rarely the cheapest. This tool compares them on your own workload so the decision is data-driven.
Start here: Self-Hosted vs API Break-Even Calculator
Results update automatically as you type.
Result
2,325,581 requests/month break-even
Self-host pays off above this volume
- API cost / request
- $0.000645
- GPU host / month
- $1,500.00
- Break-even volume
- 2,325,581 req/mo
Then: LLM Model Cost Comparison
Used unless you set active users below.
Results update automatically as you type.
Result
Cheapest: Gemma 3 4B IT at $8.00/month
Workload: 1,000 in / 500 out × 100,000 requests/month
| Gemma 3 4B IT | AWS Bedrock | $0.0000800 | $8.00 | $96.00 | 128,000 |
| Google Gemma 3 12B | AWS Bedrock | $0.000100 | $10.00 | $120.00 | 131,072 |
| Amazon Nova Micro | AWS Bedrock | $0.000105 | $10.50 | $126.00 | 128,000 |
| Command R7B | Cohere | $0.000112 | $11.25 | $135.00 | 128,000 |
| Phi 4 | Microsoft | $0.000140 | $14.00 | $168.00 | 16,384 |
| Nemotron 3 Nano 30B A3B | Nvidia | $0.000150 | $15.00 | $180.00 | 262,144 |
| Ministral 3 3B | AWS Bedrock | $0.000150 | $15.00 | $180.00 | 256,000 |
| Qwen Turbo | Alibaba | $0.000150 | $15.00 | $180.00 | 1,000,000 |
| NVIDIA Nemotron Nano 9B v2 | AWS Bedrock | $0.000175 | $17.50 | $210.00 | 128,000 |
| Amazon Nova Lite | AWS Bedrock | $0.000180 | $18.00 | $216.00 | 300,000 |
| NVIDIA Nemotron Nano 3 30B | AWS Bedrock | $0.000180 | $18.00 | $216.00 | 128,000 |
| Qwen-Omni Turbo | Alibaba | $0.000205 | $20.50 | $246.00 | 32,768 |
| gpt-oss-20b | AWS Bedrock | $0.000220 | $22.00 | $264.00 | 128,000 |
| Google Gemma 3 27B Instruct | AWS Bedrock | $0.000220 | $22.00 | $264.00 | 202,752 |
| Gemini 2.0 Flash-Lite | $0.000225 | $22.50 | $270.00 | 1,048,576 | |
| Ministral 3 8B | AWS Bedrock | $0.000225 | $22.50 | $270.00 | 128,000 |
| Mistral Nemo | Mistral AI | $0.000225 | $22.50 | $270.00 | 128,000 |
| Open Mistral Nemo | Mistral AI | $0.000225 | $22.50 | $270.00 | 128,000 |
| Pixtral 12B | Mistral AI | $0.000225 | $22.50 | $270.00 | 128,000 |
| GPT-5 Nano | OpenAI | $0.000250 | $25.00 | $300.00 | 400,000 |
| Qwen Flash | Alibaba | $0.000250 | $25.00 | $300.00 | 1,000,000 |
| Step 3.5 Flash | StepFun | $0.000250 | $25.00 | $300.00 | 256,000 |
| GLM-4.7-Flash | Z.ai (Zhipu AI) | $0.000260 | $26.00 | $312.00 | 200,000 |
| GLM-4.7-FlashX | Z.ai (Zhipu AI) | $0.000270 | $27.00 | $324.00 | 200,000 |
| GLM-4.7-Flash | AWS Bedrock | $0.000270 | $27.00 | $324.00 | 200,000 |
| Llama 4 Scout | Meta | $0.000280 | $28.00 | $336.00 | 10,000,000 |
| DeepSeek V4 Flash | DeepSeek | $0.000280 | $28.00 | $336.00 | 1,000,000 |
| MiMo-V2.5 | Xiaomi | $0.000280 | $28.00 | $336.00 | 1,048,576 |
| DeepSeek Chat | DeepSeek | $0.000280 | $28.00 | $336.00 | 1,000,000 |
| DeepSeek Reasoner | DeepSeek | $0.000280 | $28.00 | $336.00 | 1,000,000 |
| MiMo-V2-Flash | Xiaomi | $0.000280 | $28.00 | $336.00 | 262,144 |
| MiMo-V2-Omni | Xiaomi | $0.000280 | $28.00 | $336.00 | 262,144 |
| GPT-4.1 Nano | OpenAI | $0.000300 | $30.00 | $360.00 | 1,000,000 |
| Gemini 2.5 Flash-Lite | $0.000300 | $30.00 | $360.00 | 1,000,000 | |
| Gemini 2.0 Flash | $0.000300 | $30.00 | $360.00 | 1,048,576 | |
| Ministral 14B 3.0 | AWS Bedrock | $0.000300 | $30.00 | $360.00 | 128,000 |
| Qwen2.5-Omni 7B | Alibaba | $0.000300 | $30.00 | $360.00 | 32,768 |
| Llama-3.3-70B-Instruct | Meta | $0.000330 | $33.00 | $396.00 | 128,000 |
| Mistral 7B | Mistral AI | $0.000375 | $37.50 | $450.00 | 8,000 |
| Qwen-MT Turbo | Alibaba | $0.000405 | $40.50 | $486.00 | 16,384 |
| Hunyuan A13B Instruct | Tencent | $0.000425 | $42.50 | $510.00 | 131,072 |
| Nemotron 3 Super 120B A12B | Nvidia | $0.000438 | $43.75 | $525.00 | 262,144 |
| GPT-4o mini | OpenAI | $0.000450 | $45.00 | $540.00 | 128,000 |
| Mistral Small 4 | Mistral AI | $0.000450 | $45.00 | $540.00 | 131,000 |
| Command R | Cohere | $0.000450 | $45.00 | $540.00 | 128,000 |
| gpt-oss-120b | AWS Bedrock | $0.000450 | $45.00 | $540.00 | 128,000 |
| NVIDIA Nemotron 3 Super 120B A12B | AWS Bedrock | $0.000475 | $47.50 | $570.00 | 262,144 |
| DeepSeek V3.2 | DeepSeek | $0.000490 | $49.00 | $588.00 | 164,000 |
| NVIDIA Nemotron Nano 12B v2 VL BF16 | AWS Bedrock | $0.000500 | $50.00 | $600.00 | 128,000 |
| Qwen2.5 7B Instruct | Alibaba | $0.000525 | $52.50 | $630.00 | 131,072 |
| Qwen-VL Plus | Alibaba | $0.000525 | $52.50 | $630.00 | 131,072 |
| Qwen3 8B | Alibaba | $0.000530 | $53.00 | $636.00 | 131,072 |
| Hunyuan Hy3 | Tencent | $0.000600 | $60.00 | $720.00 | 262,144 |
| Llama-4-Maverick-17B-128E-Instruct-FP8 | Meta | $0.000600 | $60.00 | $720.00 | 128,000 |
| Qwen3-VL 30B-A3B | Alibaba | $0.000600 | $60.00 | $720.00 | 131,072 |
| Llama 4 Maverick | Meta | $0.000645 | $64.50 | $774.00 | 1,000,000 |
| Step 3.7 Flash | StepFun | $0.000740 | $74.00 | $888.00 | 256,000 |
| Qwen3.6 Flash | Alibaba | $0.000750 | $75.00 | $900.00 | 1,000,000 |
| GLM-4.5-Air | Z.ai (Zhipu AI) | $0.000750 | $75.00 | $900.00 | 131,072 |
| GLM-4.6V | Z.ai (Zhipu AI) | $0.000750 | $75.00 | $900.00 | 128,000 |
| GPT-5.4 nano | OpenAI | $0.000825 | $82.50 | $990.00 | 400,000 |
| Qwen/Qwen3-Next-80B-A3B-Instruct | AWS Bedrock | $0.000840 | $84.00 | $1,008.00 | 262,000 |
| Devstral 2 | Mistral AI | $0.000850 | $85.00 | $1,020.00 | 262,000 |
| DeepSeek V4 Pro | DeepSeek | $0.000870 | $87.00 | $1,044.00 | 1,000,000 |
| MiMo-V2.5-Pro | Xiaomi | $0.000870 | $87.00 | $1,044.00 | 1,048,576 |
| MiMo-V2-Pro | Xiaomi | $0.000870 | $87.00 | $1,044.00 | 1,048,576 |
| Qwen2.5-VL 7B Instruct | Alibaba | $0.000875 | $87.50 | $1,050.00 | 131,072 |
| Llama 3.3 70B | Meta | $0.000900 | $90.00 | $1,080.00 | 128,000 |
| MiniMax-M3 | MiniMax | $0.000900 | $90.00 | $1,080.00 | 1,000,000 |
| MiniMax-M2.7 | MiniMax | $0.000900 | $90.00 | $1,080.00 | 204,800 |
| MiniMax-M2 | MiniMax | $0.000900 | $90.00 | $1,080.00 | 196,608 |
| MiniMax-M2.1 | MiniMax | $0.000900 | $90.00 | $1,080.00 | 204,800 |
| MiniMax-M2.5 | MiniMax | $0.000900 | $90.00 | $1,080.00 | 204,800 |
| MiniMax M2 | AWS Bedrock | $0.000900 | $90.00 | $1,080.00 | 204,608 |
| MiniMax M2.1 | AWS Bedrock | $0.000900 | $90.00 | $1,080.00 | 204,800 |
| MiniMax M2.5 | AWS Bedrock | $0.000900 | $90.00 | $1,080.00 | 196,608 |
| Qwen3.6 35B-A3B | Alibaba | $0.000991 | $99.05 | $1,188.60 | 262,144 |
| Gemini 3.1 Flash Lite | $0.00100 | $100.00 | $1,200.00 | 1,048,576 | |
| Qwen Plus | Alibaba | $0.00100 | $100.00 | $1,200.00 | 1,000,000 |
| Qwen3-VL Plus | Alibaba | $0.00100 | $100.00 | $1,200.00 | 262,144 |
| Mixtral 8x7B | Mistral AI | $0.00105 | $105.00 | $1,260.00 | 32,000 |
| Qwen/Qwen3-VL-235B-A22B-Instruct | AWS Bedrock | $0.00105 | $105.00 | $1,260.00 | 262,000 |
| Qwen2.5 14B Instruct | Alibaba | $0.00105 | $105.00 | $1,260.00 | 131,072 |
| Qwen3 14B | Alibaba | $0.00105 | $105.00 | $1,260.00 | 131,072 |
| Qwen3 Coder Flash | Alibaba | $0.00105 | $105.00 | $1,260.00 | 1,000,000 |
| Qwen3 Coder Next | AWS Bedrock | $0.00112 | $112.00 | $1,344.00 | 131,072 |
| GPT-4.1 Mini | OpenAI | $0.00120 | $120.00 | $1,440.00 | 1,000,000 |
| Qwen Plus Character (Japanese) | Alibaba | $0.00120 | $120.00 | $1,440.00 | 8,192 |
| GPT-5 Mini | OpenAI | $0.00125 | $125.00 | $1,500.00 | 128,000 |
| Mistral Large 3 | Mistral AI | $0.00125 | $125.00 | $1,500.00 | 262,000 |
| GPT-3.5-turbo | OpenAI | $0.00125 | $125.00 | $1,500.00 | 16,385 |
| GPT-5.1 Codex mini | OpenAI | $0.00125 | $125.00 | $1,500.00 | 400,000 |
| Magistral Small | Mistral AI | $0.00125 | $125.00 | $1,500.00 | 128,000 |
| Mistral Large 3 | AWS Bedrock | $0.00125 | $125.00 | $1,500.00 | 256,000 |
| Qwen3.5 35B-A3B | Alibaba | $0.00125 | $125.00 | $1,500.00 | 262,144 |
| Qwen3-Omni Flash | Alibaba | $0.00126 | $126.00 | $1,512.00 | 65,536 |
| Mistral Medium 3.1 | Mistral AI | $0.00140 | $140.00 | $1,680.00 | 131,000 |
| Devstral 2 123B | AWS Bedrock | $0.00140 | $140.00 | $1,680.00 | 256,000 |
| Qwen3.5 27B | Alibaba | $0.00150 | $150.00 | $1,800.00 | 262,144 |
| Sonar | Perplexity | $0.00150 | $150.00 | $1,800.00 | 128,000 |
| GLM-4.5V | Z.ai (Zhipu AI) | $0.00150 | $150.00 | $1,800.00 | 64,000 |
| Qwen3-Next 80B-A3B Instruct | Alibaba | $0.00150 | $150.00 | $1,800.00 | 131,072 |
| Gemini 2.5 Flash | $0.00155 | $155.00 | $1,860.00 | 1,000,000 | |
| Gemini 3.5 Flash Lite | $0.00155 | $155.00 | $1,860.00 | 1,048,576 | |
| Qwen3-Coder 30B-A3B Instruct | Alibaba | $0.00158 | $157.50 | $1,890.00 | 262,144 |
| Qwen3.5 Plus | Alibaba | $0.00160 | $160.00 | $1,920.00 | 1,000,000 |
| GLM-4.5 | Z.ai (Zhipu AI) | $0.00170 | $170.00 | $2,040.00 | 131,072 |
| GLM-4.6 | Z.ai (Zhipu AI) | $0.00170 | $170.00 | $2,040.00 | 204,800 |
| GLM-4.7 | Z.ai (Zhipu AI) | $0.00170 | $170.00 | $2,040.00 | 204,800 |
| GLM-4.7 | AWS Bedrock | $0.00170 | $170.00 | $2,040.00 | 204,800 |
| MiniMax-M2.5-highspeed | MiniMax | $0.00180 | $180.00 | $2,160.00 | 204,800 |
| MiniMax-M2.7-highspeed | MiniMax | $0.00180 | $180.00 | $2,160.00 | 204,800 |
| Kimi K2 Thinking | Moonshot AI | $0.00185 | $185.00 | $2,220.00 | 262,144 |
| Kimi K2 Thinking | AWS Bedrock | $0.00185 | $185.00 | $2,220.00 | 262,143 |
| Qwen3.6 Plus | Alibaba | $0.00200 | $200.00 | $2,400.00 | 1,000,000 |
| Qwen3.5 122B-A10B | Alibaba | $0.00200 | $200.00 | $2,400.00 | 262,144 |
| Qwen3.7 Plus | Alibaba | $0.00200 | $200.00 | $2,400.00 | 1,000,000 |
| QwQ Plus | Alibaba | $0.00200 | $200.00 | $2,400.00 | 131,072 |
| Kimi K2.5 | Moonshot AI | $0.00210 | $210.00 | $2,520.00 | 262,144 |
| Kimi K2.5 | AWS Bedrock | $0.00210 | $210.00 | $2,520.00 | 262,143 |
| Qwen2.5 32B Instruct | Alibaba | $0.00210 | $210.00 | $2,520.00 | 131,072 |
| Qwen3 235B-A22B | Alibaba | $0.00210 | $210.00 | $2,520.00 | 131,072 |
| Qwen3 32B | Alibaba | $0.00210 | $210.00 | $2,520.00 | 131,072 |
| Qwen3-VL 235B-A22B | Alibaba | $0.00210 | $210.00 | $2,520.00 | 131,072 |
| Amazon Nova Pro | AWS Bedrock | $0.00240 | $240.00 | $2,880.00 | 300,000 |
| Nemotron 3 Ultra 550B A55B | Nvidia | $0.00240 | $240.00 | $2,880.00 | 1,000,000 |
| Qwen-VL Max | Alibaba | $0.00240 | $240.00 | $2,880.00 | 131,072 |
| Qwen3.5 397B-A17B | Alibaba | $0.00240 | $240.00 | $2,880.00 | 262,144 |
| Qwen3.6 27B | Alibaba | $0.00240 | $240.00 | $2,880.00 | 262,144 |
| Grok 4.3 | xAI | $0.00250 | $250.00 | $3,000.00 | 1,000,000 |
| Grok 4.3 | AWS Bedrock | $0.00250 | $250.00 | $3,000.00 | 1,000,000 |
| GLM-5 | Z.ai (Zhipu AI) | $0.00260 | $260.00 | $3,120.00 | 204,800 |
| GLM-5 | AWS Bedrock | $0.00260 | $260.00 | $3,120.00 | 202,752 |
| MiMo-V2.5-Pro-UltraSpeed | Xiaomi | $0.00261 | $261.00 | $3,132.00 | 1,048,576 |
| Kimi K2.7 Code | Moonshot AI | $0.00295 | $295.00 | $3,540.00 | 262,144 |
| Kimi K2.6 | Moonshot AI | $0.00295 | $295.00 | $3,540.00 | 262,144 |
| MAI-Code-1-Flash | Microsoft | $0.00300 | $300.00 | $3,600.00 | 256,000 |
| GPT-5.4 mini | OpenAI | $0.00300 | $300.00 | $3,600.00 | 400,000 |
| GLM-5-Turbo | Z.ai (Zhipu AI) | $0.00320 | $320.00 | $3,840.00 | 200,000 |
| GLM-5V-Turbo | Z.ai (Zhipu AI) | $0.00320 | $320.00 | $3,840.00 | 200,000 |
| o4-mini | OpenAI | $0.00330 | $330.00 | $3,960.00 | 200,000 |
| o3-mini | OpenAI | $0.00330 | $330.00 | $3,960.00 | 200,000 |
| Muse Spark 1.1 | Meta | $0.00338 | $337.50 | $4,050.00 | 1,000,000 |
| Claude Haiku 4.5 | Anthropic | $0.00350 | $350.00 | $4,200.00 | 200,000 |
| Qwen3 Coder Plus | Alibaba | $0.00350 | $350.00 | $4,200.00 | 1,048,576 |
| Qwen3-Next 80B-A3B (Thinking) | Alibaba | $0.00350 | $350.00 | $4,200.00 | 131,072 |
| GLM-5.2 | Z.ai (Zhipu AI) | $0.00360 | $360.00 | $4,320.00 | 1,000,000 |
| GLM-5.1 | Z.ai (Zhipu AI) | $0.00360 | $360.00 | $4,320.00 | 200,000 |
| QVQ Max | Alibaba | $0.00360 | $360.00 | $4,320.00 | 131,072 |
| GPT-5.6 Luna | OpenAI | $0.00400 | $400.00 | $4,800.00 | 1,050,000 |
| GPT-5.6 Luna | AWS Bedrock | $0.00400 | $400.00 | $4,800.00 | 272,000 |
| Qwen3 Max | Alibaba | $0.00420 | $420.00 | $5,040.00 | 262,144 |
| Qwen2.5 72B Instruct | Alibaba | $0.00420 | $420.00 | $5,040.00 | 131,072 |
| Qwen Max | Alibaba | $0.00480 | $480.00 | $5,760.00 | 32,768 |
| Grok 4.5 | xAI | $0.00500 | $500.00 | $6,000.00 | 500,000 |
| Mixtral 8x22B | Mistral AI | $0.00500 | $500.00 | $6,000.00 | 64,000 |
| Kimi K2 Thinking Turbo | Moonshot AI | $0.00515 | $515.00 | $6,180.00 | 262,144 |
| Qwen3-Coder 480B-A35B Instruct | Alibaba | $0.00525 | $525.00 | $6,300.00 | 262,144 |
| Gemini 3.6 Flash | $0.00525 | $525.00 | $6,300.00 | 1,048,576 | |
| Kimi K2.7 Code HighSpeed | Moonshot AI | $0.00590 | $590.00 | $7,080.00 | 262,144 |
| GPT-4.1 | OpenAI | $0.00600 | $600.00 | $7,200.00 | 1,000,000 |
| o3 | OpenAI | $0.00600 | $600.00 | $7,200.00 | 200,000 |
| Sonar Reasoning Pro | Perplexity | $0.00600 | $600.00 | $7,200.00 | 128,000 |
| Sonar Deep Research | Perplexity | $0.00600 | $600.00 | $7,200.00 | 128,000 |
| Gemini 3.5 Flash | $0.00600 | $600.00 | $7,200.00 | 1,048,576 | |
| o4-mini-deep-research | OpenAI | $0.00600 | $600.00 | $7,200.00 | 200,000 |
| Qwen-MT Plus | Alibaba | $0.00615 | $614.50 | $7,374.00 | 16,384 |
| GPT-5 | OpenAI | $0.00625 | $625.00 | $7,500.00 | 128,000 |
| Gemini 2.5 Pro | $0.00625 | $625.00 | $7,500.00 | 1,048,576 | |
| GPT-5.1 | OpenAI | $0.00625 | $625.00 | $7,500.00 | 400,000 |
| GPT-5.1 Codex | OpenAI | $0.00625 | $625.00 | $7,500.00 | 400,000 |
| GPT-5.1 Codex Max | OpenAI | $0.00625 | $625.00 | $7,500.00 | 400,000 |
| GPT-5-Codex | OpenAI | $0.00625 | $625.00 | $7,500.00 | 400,000 |
| Qwen3.7 Max | Alibaba | $0.00625 | $625.00 | $7,500.00 | 1,000,000 |
| Step 1 (32K) | StepFun | $0.00684 | $684.50 | $8,214.00 | 32,768 |
| Claude Sonnet 5 | Anthropic | $0.00700 | $700.00 | $8,400.00 | 1,000,000 |
| Claude Sonnet 5 (Global) | AWS Bedrock | $0.00700 | $700.00 | $8,400.00 | 1,000,000 |
| Qwen2.5-VL 72B Instruct | Alibaba | $0.00700 | $700.00 | $8,400.00 | 131,072 |
| GPT-4o | OpenAI | $0.00750 | $750.00 | $9,000.00 | 128,000 |
| Command A | Cohere | $0.00750 | $750.00 | $9,000.00 | 256,000 |
| Command R+ | Cohere | $0.00750 | $750.00 | $9,000.00 | 128,000 |
| Amazon Nova Premier | AWS Bedrock | $0.00875 | $875.00 | $10,500.00 | 1,000,000 |
| GPT-5.2 | OpenAI | $0.00875 | $875.00 | $10,500.00 | 400,000 |
| GPT-5.2 Codex | OpenAI | $0.00875 | $875.00 | $10,500.00 | 400,000 |
| GPT-5.3 Codex | OpenAI | $0.00875 | $875.00 | $10,500.00 | 400,000 |
| GPT-5.3 Codex Spark | OpenAI | $0.00875 | $875.00 | $10,500.00 | 128,000 |
| GPT-5.4 | OpenAI | $0.0100 | $1,000.00 | $12,000.00 | 1,050,000 |
| GPT-5.6 Terra | OpenAI | $0.0100 | $1,000.00 | $12,000.00 | 1,050,000 |
| GPT-5.6 Terra | AWS Bedrock | $0.0100 | $1,000.00 | $12,000.00 | 272,000 |
| Claude Sonnet 4.6 | Anthropic | $0.0105 | $1,050.00 | $12,600.00 | 1,000,000 |
| Sonar Pro | Perplexity | $0.0105 | $1,050.00 | $12,600.00 | 200,000 |
| Kimi K3 | Moonshot AI | $0.0105 | $1,050.00 | $12,600.00 | 1,048,576 |
| Claude Sonnet 4.5 (latest) | Anthropic | $0.0105 | $1,050.00 | $12,600.00 | 1,000,000 |
| GPT-5.4 | AWS Bedrock | $0.0110 | $1,100.00 | $13,200.00 | 272,000 |
| AU Anthropic Claude Sonnet 4.6 | AWS Bedrock | $0.0116 | $1,155.00 | $13,860.00 | 1,000,000 |
| Step 2 (16K) | StepFun | $0.0134 | $1,343.00 | $16,116.00 | 16,384 |
| Claude Opus 4.8 | Anthropic | $0.0175 | $1,750.00 | $21,000.00 | 1,000,000 |
| Claude Opus 4.6 | Anthropic | $0.0175 | $1,750.00 | $21,000.00 | 1,000,000 |
| Claude Opus 4.6 (US) | AWS Bedrock | $0.0175 | $1,750.00 | $21,000.00 | 1,000,000 |
| Claude Opus 4.7 (US) | AWS Bedrock | $0.0175 | $1,750.00 | $21,000.00 | 1,000,000 |
| Claude Opus 4.8 | AWS Bedrock | $0.0175 | $1,750.00 | $21,000.00 | 1,000,000 |
| Claude Opus 4.5 (latest) | Anthropic | $0.0175 | $1,750.00 | $21,000.00 | 200,000 |
| Claude Opus 4.7 | Anthropic | $0.0175 | $1,750.00 | $21,000.00 | 1,000,000 |
| Claude Opus 5 (JP) | AWS Bedrock | $0.0175 | $1,750.00 | $21,000.00 | 1,000,000 |
| Claude Opus 5 | Anthropic | $0.0175 | $1,750.00 | $21,000.00 | 1,000,000 |
| GPT-5.5 | OpenAI | $0.0200 | $2,000.00 | $24,000.00 | 1,050,000 |
| GPT-5.6 | OpenAI | $0.0200 | $2,000.00 | $24,000.00 | 1,050,000 |
| GPT-5.6 Sol | OpenAI | $0.0200 | $2,000.00 | $24,000.00 | 1,050,000 |
| GPT-5.6 Sol | AWS Bedrock | $0.0200 | $2,000.00 | $24,000.00 | 272,000 |
| GPT-5.5 | AWS Bedrock | $0.0220 | $2,200.00 | $26,400.00 | 272,000 |
| GPT-4 Turbo | OpenAI | $0.0250 | $2,500.00 | $30,000.00 | 128,000 |
| o3-deep-research | OpenAI | $0.0300 | $3,000.00 | $36,000.00 | 200,000 |
| Claude Fable 5 | Anthropic | $0.0350 | $3,500.00 | $42,000.00 | 1,000,000 |
| Claude Fable 5 | AWS Bedrock | $0.0350 | $3,500.00 | $42,000.00 | 1,000,000 |
| o1 | OpenAI | $0.0450 | $4,500.00 | $54,000.00 | 200,000 |
| Claude Opus 4.1 (latest) | Anthropic | $0.0525 | $5,250.00 | $63,000.00 | 200,000 |
| GPT-4 | OpenAI | $0.0600 | $6,000.00 | $72,000.00 | 8,192 |
| o3-pro | OpenAI | $0.0600 | $6,000.00 | $72,000.00 | 200,000 |
| GPT-5 Pro | OpenAI | $0.0750 | $7,500.00 | $90,000.00 | 400,000 |
| GPT-5.2 Pro | OpenAI | $0.1050 | $10,500.00 | $126,000.00 | 400,000 |
| GPT-5.4 Pro | OpenAI | $0.1200 | $12,000.00 | $144,000.00 | 1,050,000 |
| GPT-5.5 Pro | OpenAI | $0.1200 | $12,000.00 | $144,000.00 | 1,050,000 |
| o1-pro | OpenAI | $0.4500 | $45,000.00 | $540,000.00 | 200,000 |
Why this isn't trivial
The part people underestimate: open models trade per-token API fees for fixed GPU and ops costs, so the winner depends heavily on your utilization. In practice the biggest savings come from self-hosting only when utilization is high enough to beat per-token pricing, so it is worth modelling before you commit.
How it's calculated
We estimate this by comparing per-token API cost at your volume against amortized GPU hosting for an open model. Every figure uses the current provider prices baked into the site (reviewed daily), and you can override any input to match your own assumptions.
Frequently asked questions
When does self-hosting win?+
At high, steady utilization where fixed GPU cost spreads over many requests.
What hidden costs does open have?+
Ops, scaling and reliability engineering — include them in the comparison.
Are these prices up to date?+
Yes. The model prices behind this calculator are refreshed and reviewed daily, so your estimate reflects current provider rates rather than a stale snapshot.
Related
Should I self-host an LLM or keep using an API?
Decide whether self-hosting an open LLM beats paying per-token API fees at your volume.
At what scale does self-hosting an LLM pay off?
Find the request volume at which self-hosting an LLM beats per-token API pricing.
GPT-5 vs Claude: which is cheaper for a support bot?
Compare GPT-5 and Claude on the real cost of running a support bot at your volume using today's prices.
Gemini vs GPT: what's the annual cost difference?
See the annual cost gap between Gemini and GPT for your workload, projected from per-request cost.
Calculator
Self-Hosted vs API Break-Even Calculator
Calculator
LLM Model Cost Comparison
Estimates for planning. Pricing data last reviewed 28 July 2026.