is now an ernd partner: 10% back on inference On this page
Cheapest and fastest
- Cheapest
- Baidu Qianfan
- $0.0667 per 1M tokens blended
- $0.0444 input, $0.133 output
- Fastest
- Cohere
- 166 tokens per second
- $0.175 blended, 0.35 s to first token
Every provider
| Provider | Input /1M | Output /1M | Blended 3:1 | Cache read /1M | Context | Quantization | Latency | Throughput | With ernd card |
|---|---|---|---|---|---|---|---|---|---|
| Baidu Qianfan | $0.0444 | $0.133 | $0.0667 | $0.0014 | 1.05M | FP8 | 0.86 s | 96 tok/s | – |
| DeepInfra | $0.06 | $0.18 | $0.09 | $0.015 | 1.05M | FP8 | 0.80 s | 46 tok/s | – |
| Wafer (US) ernd partner | $0.0434 | $0.30 | $0.108 | $0.035 | 1.05M | NVFP4 | 0.34 s | 140 tok/s | $0.0968after 10% cash back |
| Prime Intellect | $0.09 | $0.23 | $0.125 | $0.029 | 1.05M | FP4 | 0.62 s | 94 tok/s | – |
| StreamLake | $0.088 | $0.264 | $0.132 | $0.0028 | 1.02M | FP8 | 2.00 s | 39 tok/s | – |
| Wafer (Fast) ernd partner | $0.10 | $0.23 | $0.133 | $0.095 | 1.05M | NVFP4 | 0.30 s | 141 tok/s | $0.119after 10% cash back |
| Reka AI | $0.10 | $0.28 | $0.145 | $0.04 | 262K | Not listed | 0.64 s | 101 tok/s | – |
| DigitalOcean | $0.119 | $0.238 | $0.149 | $0.0238 | 1.05M | Not listed | 0.68 s | 140 tok/s | – |
| Baseten (fp32) | $0.13 | $0.26 | $0.163 | $0.028 | 1.05M | FP32 | 0.30 s | 109 tok/s | – |
| Baseten (fp8) | $0.13 | $0.26 | $0.163 | $0.028 | 1.05M | FP8 | 0.33 s | 137 tok/s | – |
| Venice | $0.13 | $0.26 | $0.163 | $0.028 | 1M | Not listed | 0.59 s | 114 tok/s | – |
| CoreWeave | $0.13 | $0.27 | $0.165 | $0.05 | 262K | FP8 | 0.42 s | 103 tok/s | – |
| Cohere | $0.14 | $0.28 | $0.175 | $0.07 | 1.05M | FP8 | 0.35 s | 166 tok/s | – |
| Nebius Token Factory | $0.14 | $0.28 | $0.175 | $0.01 | 1.05M | FP8 | 0.95 s | 63 tok/s | – |
| Parasail | $0.14 | $0.28 | $0.175 | $0.05 | 1.05M | FP8 | 0.62 s | 83 tok/s | – |
| Sail Research | $0.15 | $0.30 | $0.188 | $0.042 | 1.05M | FP4 | 0.98 s | 47 tok/s | – |
| Sail Research (US) | $0.15 | $0.30 | $0.188 | $0.042 | 1.05M | FP4 | 0.76 s | 43 tok/s | – |
| Morph | $0.142 | $0.40 | $0.206 | $0.036 | 1.05M | BF16 | 0.61 s | 96 tok/s | – |
| Inceptron | $0.13 | $0.495 | $0.221 | $0.027 | 1.05M | FP4 | 0.82 s | 54 tok/s | – |
| Mancer | $0.175 | $0.50 | $0.256 | $0.175 | 1.05M | FP8 | 0.95 s | 56 tok/s | – |
| Alibaba Cloud Int. | $0.176 | $0.528 | $0.264 | $0.0176 | 1M | Not listed | 1.86 s | 96 tok/s | – |
| Alibaba Cloud Int. (US-EAST-1) | $0.187 | $0.561 | $0.28 | $0.0187 | 1M | Not listed | 1.83 s | 98 tok/s | – |
| SiliconFlow | $0.22 | $0.66 | $0.33 | $0.028 | 1.05M | FP8 | 1.45 s | 32 tok/s | – |
| GMICloud | $0.286 | $0.858 | $0.429 | $0.0091 | 1.05M | FP8 | 2.00 s | 60 tok/s | – |
| Phala | $0.308 | $0.924 | $0.462 | $0.0196 | 1.05M | Not listed | 1.80 s | 100 tok/s | – |
| NovitaAI | $0.409 | $1.23 | $0.614 | $0.026 | 1.05M | FP8 | 1.07 s | 60 tok/s | – |
| AtlasCloud | $0.44 | $1.32 | $0.66 | $0.028 | 1.05M | FP4 | 1.61 s | 77 tok/s | – |
| Chutes | $0.44 | $1.32 | $0.66 | $0.044 | 1.05M | FP8 | 1.76 s | 64 tok/s | – |
| Cloudflare | $0.44 | $1.32 | $0.66 | $0.014 | 1.05M | Not listed | 1.32 s | 68 tok/s | – |
Blended is three parts input price to one part output, per million tokens. Latency is time to first token and throughput is output tokens per second, both averaged from OpenRouter's daily figures for the last 7 days. Cheapest and fastest only count listings that were up when checked.
Wafer is an ernd partner: 10% back on Wafer inference with the ernd card. With ernd card is the blended price less 10% cash back on Wafer charges. It comes back on the card bill and doesn't change Wafer's list price. It applies when you pay Wafer directly, not through OpenRouter credits.
Questions
What is the cheapest DeepSeek V4 Flash 0731 API provider?
Baidu Qianfan, at $0.0667 per million tokens blended ($0.0444 input, $0.133 output), in OpenRouter's listings on November 7, 2026.
How much does DeepSeek V4 Flash 0731 cost per million tokens?
Input runs from $0.0434 to $0.44 per million tokens across 29 listings, and output from $0.133 to $1.32. The middle listing comes to $0.175 blended.
Which DeepSeek V4 Flash 0731 provider is fastest?
Cohere had the most throughput over the last 7 days, about 166 tokens per second. Baseten (fp32) was quickest to the first token at 0.30 s.
How do I pay for DeepSeek V4 Flash 0731 inference?
Buy OpenRouter credits and it sends each request to one of these providers, or open an account with a provider and pay it directly. The ernd card is a credit card for US companies. It earns 4% back on AI vendor spend and 1.5% back on everything else. Wafer charges on the card earn 10% back.
More models
- DeepSeek V4.1 Flash 1017 from $0.0813 blended
- DeepSeek V4.1 Flash from $0.0887 blended
- GLM 5.3 Flash from $0.0459 blended
- Hy4 preview from $1.13 blended
- MiniMax M3.1 Flash from $0.131 blended
- GLM 5.3 from $0.402 blended
- Kimi K3.1 Pro from $2.27 blended
- Nemotron 3 Ultra from $0.74 blended
- Hy3 from $0.144 blended
- Nemotron 3.5 Super VL from $0.133 blended
- MiMo-V2.6-Flash from $0.14 blended
- DeepSeek V4 Flash 0423 from $0.0544 blended
- Kimi K3 from $2.61 blended
- Nemotron 3.5 Lightning from $0.0488 blended
- MiMo-V2.6-Pro from $0.486 blended
- GLM 5.2 from $0.29 blended
- DeepSeek V4 Pro 0813 from $0.871 blended
- Qwen3.8 27B from $0.342 blended
- gpt-oss-120b from $0.025 blended
Source
Checked . Providers change prices often, so OpenRouter's page is the final word. ernd is run by Provenance Research Inc.