What your workload costs on each model.
Describe a typical request and how many you send. Every model’s monthly cost at its listed price, next to its capability index.
Your workload
Pick a typical request or enter your own numbers. The link keeps them, to share.
Monthly cost and capability
Each model on the capability index, at its monthly cost for your workload. Up and to the left does more for less; the line marks the highest index at each cost.
OpenAI
Alibaba
Google
Anthropic
DeepSeek
Mistral AI
Meta
xAI
Moonshot AI
Z.ai
MiniMax
Microsoft
NVIDIA- Highest index at each cost
Use the arrow keys to move between models. Press Enter to open the selected model. A table with the same data follows the chart.
Every model
174 models, cheapest first.
| Model | Capability index | Price per 1M, in / out | Per request | Per month |
|---|---|---|---|---|
| Llama 3 8B InstructMeta, open weights | 116.5 | $0.04 / $0.04 | < $0.01 | $0.24 |
| Qwen3.7 FlashAlibaba | 144.6 | $0.03 / $0.13 | < $0.01 | $0.33 |
| Llama-3.1-8B-InstructMeta, open weights | 116.6 | $0.06 / $0.06 | < $0.01 | $0.36 |
| Gemma 3 4B ITGoogle, open weights | 116.0 | $0.05 / $0.10 | < $0.01 | $0.38 |
| Llama 3.2 1B InstructMeta, open weights | 102.0 | $0.027 / $0.201 | < $0.01 | $0.42 |
| Gemma 3 12B ITGoogle, open weights | 123.5 | $0.05 / $0.15 | < $0.01 | $0.45 |
| GPT OSS 20BOpenAI, open weights | 137.8 | $0.05 / $0.20 | < $0.01 | $0.52 |
| nemotron-3-ultraNVIDIA, open weights | 146.2 | $0.10 / $0.10 | < $0.01 | $0.60 |
| Qwen2.5-Coder-0.5BAlibaba, open weights | 88.2 | $0.10 / $0.10 | < $0.01 | $0.60 |
| Mistral Nemo Instruct 2407Mistral AI, open weights | 118.7 | $0.10 / $0.121 | < $0.01 | $0.63 |
| Gemma 4 26B A4B ITGoogle, open weights | 141.8 | $0.05 / $0.29 | < $0.01 | $0.66 |
| Mistral-7B-Instruct-v0.3Mistral AI, open weights | 109.0 | $0.111 / $0.111 | < $0.01 | $0.67 |
| Qwen3.5 9BAlibaba, open weights | 139.5 | $0.10 / $0.15 | < $0.01 | $0.67 |
| GPT-5 NanoOpenAI | 139.4 | $0.05 / $0.40 | < $0.01 | $0.83 |
| Qwen3 30B A3B Instruct 2507Alibaba, open weights | 137.4 | $0.10 / $0.30 | < $0.01 | $0.90 |
| DeepSeek R1 Distill Qwen 14BDeepSeek, open weights | 135.4 | $0.15 / $0.15 | < $0.01 | $0.90 |
| Mistral Small 3.2Mistral AI, open weights | 131.7 | $0.10 / $0.30 | < $0.01 | $0.90 |
| Mistral Small 3.1Mistral AI | 127.5 | $0.10 / $0.30 | < $0.01 | $0.90 |
| gemini-2.5-flash-lite-preview-06-17Google | 133.9 | $0.09 / $0.36 | < $0.01 | $0.95 |
| Gemma 3 27B ITGoogle, open weights | 130.0 | $0.08 / $0.45 | < $0.01 | $1.04 |
| DeepSeek V4 Flash 0731DeepSeek, open weights | 154.5 | $0.14 / $0.28 | < $0.01 | $1.05 |
| Qwen3.5 FlashAlibaba | 144.0 | $0.10 / $0.40 | < $0.01 | $1.05 |
| GPT-4.1 nanoOpenAI | 129.6 | $0.10 / $0.40 | < $0.01 | $1.05 |
| Qwen2.5-Coder 7B InstructAlibaba, open weights | 113.1 | $0.144 / $0.287 | < $0.01 | $1.08 |
| GPT-6 LunaOpenAI | 156.3 | $0.10 / $0.50 | < $0.01 | $1.20 |
Calculated by Artificials. Each model is at its listed price per 1M tokens: the maker’s own paid listing, or else the middle paid listing across hosts, from models.dev (MIT). Capability index from Epoch AI (CC BY 4.0). A month is 30 days. Cached input is charged at a model’s cached price where it lists one, and at its full price where it does not. Batch discounts, free tiers, taxes and minimum fees are not included, and a model that thinks before it answers can write more tokens than you set.