Price per 1M tokens using the Artificial Analysis 3:1 blended ratio.
Blended Price is a cost metric for comparing providers with different input and output token prices. Artificial Analysis calculates it with a 3:1 input-to-output token ratio so a single number can summarize typical API cost.
Test type: Provider price normalization. Lower values rank better.
325 models have this metric.
Current leader: Qwen3.5 0.8B (Non-reasoning)
Project links
Prices come from Artificial Analysis pricing data in the committed snapshot.
Top models ranked by Blended $.
| Rank | Model | Creator | Value | Speed | Blended Price |
|---|---|---|---|---|---|
| #1 | Qwen3.5 0.8B (Non-reasoning) | Alibaba | $0.020/M | 273.6 tok/s | $0.020/M |
| #2 |
| Alibaba |
| $0.020/M |
| n/a |
| $0.020/M |
| #3 | Gemma 3n E4B Instruct | $0.025/M | 15.3 tok/s | $0.025/M |
| #4 | Qwen3.5 2B (Non-reasoning) | Alibaba | $0.040/M | 227 tok/s | $0.040/M |
| #5 | Qwen3.5 2B (Reasoning) | Alibaba | $0.040/M | n/a | $0.040/M |
| #6 | LFM2 24B A2B | Liquid AI | $0.052/M | 196.9 tok/s | $0.052/M |
| #7 | Qwen3.5 4B (Non-reasoning) | Alibaba | $0.060/M | 200.2 tok/s | $0.060/M |
| #8 | Qwen3.5 4B (Reasoning) | Alibaba | $0.060/M | 204.8 tok/s | $0.060/M |
| #9 | Nova Micro | Amazon | $0.061/M | 332.1 tok/s | $0.061/M |
| #10 | Granite 4.1 8B | IBM | $0.063/M | 134.6 tok/s | $0.063/M |
| #11 | Llama 3 Instruct 8B | Meta | $0.070/M | 82.2 tok/s | $0.070/M |
| #12 | NVIDIA Nemotron Nano 9B V2 (Reasoning) | NVIDIA | $0.070/M | 121.6 tok/s | $0.070/M |
| #13 | Granite 3.3 8B (Non-reasoning) | IBM | $0.085/M | 410.5 tok/s | $0.085/M |
| #14 | NVIDIA Nemotron Nano 9B V2 (Non-reasoning) | NVIDIA | $0.086/M | 153.3 tok/s | $0.086/M |
| #15 | Qwen2.5 Turbo | Alibaba | $0.087/M | 77.7 tok/s | $0.087/M |
| #16 | NVIDIA Nemotron 3 Nano 30B A3B (Reasoning) | NVIDIA | $0.096/M | 154.8 tok/s | $0.096/M |
| #17 | gpt-oss-20B (high) | OpenAI | $0.100/M | 242.3 tok/s | $0.100/M |
| #18 | Llama 2 Chat 7B | Meta | $0.100/M | 99.7 tok/s | $0.100/M |
| #19 | Llama 3.1 Instruct 8B | Meta | $0.100/M | 164.4 tok/s | $0.100/M |
| #20 | Llama 3.2 Instruct 1B | Meta | $0.100/M | 97.7 tok/s | $0.100/M |
| #21 | Ministral 3 3B | Mistral | $0.100/M | 287.6 tok/s | $0.100/M |
| #22 | Nova Lite | Amazon | $0.105/M | 186.8 tok/s | $0.105/M |
| #23 | Granite 4.0 H Small | IBM | $0.107/M | 238.9 tok/s | $0.107/M |
| #24 | gpt-oss-20B (low) | OpenAI | $0.108/M | 249.7 tok/s | $0.108/M |
| #25 | Qwen3.5 9B (Reasoning) | Alibaba | $0.113/M | 62.9 tok/s | $0.113/M |
| #26 | Apertus 8B Instruct | Swiss AI Initiative | $0.125/M | n/a | $0.125/M |
| #27 | Olmo 3 7B Instruct | Allen Institute for AI | $0.125/M | n/a | $0.125/M |
| #28 | GPT-5 nano (high) | OpenAI | $0.138/M | 136 tok/s | $0.138/M |
| #29 | GPT-5 nano (medium) | OpenAI | $0.138/M | 150.3 tok/s | $0.138/M |
| #30 | GPT-5 nano (minimal) | OpenAI | $0.138/M | 139.1 tok/s | $0.138/M |
| #31 | Devstral Small (Jul '25) | Mistral | $0.150/M | 194.2 tok/s | $0.150/M |
| #32 | Ling 2.6 Flash | InclusionAI | $0.150/M | 206 tok/s | $0.150/M |
| #33 | Llama 3.2 Instruct 3B | Meta | $0.150/M | 52.2 tok/s | $0.150/M |
| #34 | MiMo-V2-Flash (Feb 2026) | Xiaomi | $0.150/M | 120.6 tok/s | $0.150/M |
| #35 | MiMo-V2-Flash (Non-reasoning) | Xiaomi | $0.150/M | 116.7 tok/s | $0.150/M |
| #36 | MiMo-V2-Flash (Reasoning) | Xiaomi | $0.150/M | 118.8 tok/s | $0.150/M |
| #37 | Ministral 3 8B | Mistral | $0.150/M | 157.6 tok/s | $0.150/M |
| #38 | Mistral Small 3 | Mistral | $0.150/M | 135.9 tok/s | $0.150/M |
| #39 | Mistral Small 3.1 | Mistral | $0.150/M | 138.8 tok/s | $0.150/M |
| #40 | Mistral Small 3.2 | Mistral | $0.150/M | 153.8 tok/s | $0.150/M |
| #41 | Solar Mini | Upstage | $0.150/M | 41.7 tok/s | $0.150/M |
| #42 | Step 3.5 Flash | StepFun | $0.150/M | 123.6 tok/s | $0.150/M |
| #43 | GLM-4.7-Flash (Non-reasoning) | Z AI | $0.152/M | 89.6 tok/s | $0.152/M |
| #44 | GLM-4.7-Flash (Reasoning) | Z AI | $0.152/M | 110.5 tok/s | $0.152/M |
| #45 | DeepSeek V4 Flash (Reasoning, High Effort) | DeepSeek | $0.175/M | n/a | $0.175/M |
| #46 | DeepSeek V4 Flash (Reasoning, Max Effort) | DeepSeek | $0.175/M | 77.4 tok/s | $0.175/M |
| #47 | Gemini 2.5 Flash-Lite (Non-reasoning) | $0.175/M | 239.9 tok/s | $0.175/M |
| #48 | Gemini 2.5 Flash-Lite (Reasoning) | $0.175/M | 243.6 tok/s | $0.175/M |
| #49 | Gemini 2.5 Flash-Lite Preview (Sep '25) (Non-reasoning) | $0.175/M | n/a | $0.175/M |
| #50 | Gemini 2.5 Flash-Lite Preview (Sep '25) (Reasoning) | $0.175/M | n/a | $0.175/M |
| #51 | GPT-4.1 nano | OpenAI | $0.175/M | 125.2 tok/s | $0.175/M |
| #52 | Llama Nemotron Super 49B v1.5 (Non-reasoning) | NVIDIA | $0.175/M | 51.3 tok/s | $0.175/M |
| #53 | Llama Nemotron Super 49B v1.5 (Reasoning) | NVIDIA | $0.175/M | 50.8 tok/s | $0.175/M |
| #54 | Qwen3 0.6B (Non-reasoning) | Alibaba | $0.188/M | 204.8 tok/s | $0.188/M |
| #55 | Qwen3 1.7B (Non-reasoning) | Alibaba | $0.188/M | 139 tok/s | $0.188/M |
| #56 | Qwen3 4B (Non-reasoning) | Alibaba | $0.188/M | 104.4 tok/s | $0.188/M |
| #57 | Gemma 4 26B A4B (Reasoning) | $0.198/M | n/a | $0.198/M |
| #58 | Hermes 4 - Llama-3.1 70B (Non-reasoning) | Nous Research | $0.198/M | 83.5 tok/s | $0.198/M |
| #59 | Hermes 4 - Llama-3.1 70B (Reasoning) | Nous Research | $0.198/M | 78.6 tok/s | $0.198/M |
| #60 | Ministral 3 14B | Mistral | $0.200/M | 121.6 tok/s | $0.200/M |
| #61 | Phi-4 | Microsoft | $0.219/M | 41.6 tok/s | $0.219/M |
| #62 | Llama 3.2 Instruct 11B (Vision) | Meta | $0.245/M | 77.4 tok/s | $0.245/M |
| #63 | Ling-flash-2.0 | InclusionAI | $0.247/M | 87.3 tok/s | $0.247/M |
| #64 | Ring-flash-2.0 | InclusionAI | $0.247/M | 91 tok/s | $0.247/M |
| #65 | Jamba 1.5 Mini | AI21 Labs | $0.250/M | n/a | $0.250/M |
| #66 | Jamba 1.6 Mini | AI21 Labs | $0.250/M | 184.5 tok/s | $0.250/M |
| #67 | Mistral 7B Instruct | Mistral | $0.250/M | 156.9 tok/s | $0.250/M |
| #68 | Gemini 2.0 Flash (Feb '25) | $0.263/M | n/a | $0.263/M |
| #69 | GPT-4o mini | OpenAI | $0.263/M | 59.9 tok/s | $0.263/M |
| #70 | gpt-oss-120B (high) | OpenAI | $0.263/M | 212.3 tok/s | $0.263/M |
| #71 | gpt-oss-120B (low) | OpenAI | $0.263/M | 216.3 tok/s | $0.263/M |
| #72 | Mistral Small 4 (Non-reasoning) | Mistral | $0.263/M | 139.5 tok/s | $0.263/M |
| #73 | Mistral Small 4 (Reasoning) | Mistral | $0.263/M | 149.5 tok/s | $0.263/M |
| #74 | Grok 4 Fast (Non-reasoning) | xAI | $0.275/M | 77.4 tok/s | $0.275/M |
| #75 | Grok 4 Fast (Reasoning) | xAI | $0.275/M | 76.2 tok/s | $0.275/M |
| #76 | Grok 4.1 Fast (Non-reasoning) | xAI | $0.275/M | 112.1 tok/s | $0.275/M |
| #77 | Grok 4.1 Fast (Reasoning) | xAI | $0.275/M | 140.9 tok/s | $0.275/M |
| #78 | Qwen3.5 Omni Flash | Alibaba | $0.275/M | 190.4 tok/s | $0.275/M |
| #79 | Llama 4 Scout | Meta | $0.292/M | 109.2 tok/s | $0.292/M |
| #80 | Hermes 3 - Llama-3.1 70B | Nous Research | $0.300/M | 28.8 tok/s | $0.300/M |