DeepInfra: Models and Prices
Serverless token API and dedicated deployments for open-weight models.
DeepInfra served 6.43T tokens of other labs' open-weight models over Oct 3 to Oct 9, 2026 through a large public LLM router, #3 among providers.
Open-weight models
| Model | Family | Share of model | Tokens, 7d | Input $/M | Output $/M | Tokens/s | Uptime 3d | Context |
|---|---|---|---|---|---|---|---|---|
| DeepSeek V4.1 Flash | DeepSeek | 14.2% | 5.34T | $0.14 | $0.42 | 49 | 100.0% | 1.05M |
| DeepSeek V4 Flash 0731 | DeepSeek | 5.8% | 295B | $0.06 | $0.18 | 43 | 100.0% | 1.05M |
| Mistral Nemo | Mistral | 91.7% | 223B | $0.029 | $0.03 | 21 | 100.0% | 131K |
| Gemma 4 31B | Gemma | 37.3% | 158B | $0.09 | $0.34 | 35 | 100.0% | 262K |
| GLM 5.3 Flash | GLM | 1.0% | 109B | $0.075 | $0.25 | 44 | 99.2% | 1.05M |
| DeepSeek V4 Flash 0423 | DeepSeek | 3.1% | 94B | $0.09 | $0.18 | 23 | 100.0% | 1.05M |
| gpt-oss-120b | gpt-oss | 22.0% | 92B | $0.037 | $0.17 | 73 | 98.5% | 131K |
| MiMo-V2.6-Pro | MiMo | 4.1% | 50B | $0.43 | $0.87 | 44 | 100.0% | 1.05M |
| MiMo-V2.6-Flash | MiMo | 0.3% | 31B | $0.14 | $0.28 | 25 | 99.6% | 1.05M |
| GLM 5.3 | GLM | 0.7% | 24B | $0.563 | $2.50 | 80 | 98.3% | 1.05M |
| DeepSeek V4 Flash Vision Exp | DeepSeek | 77.0% | 20B | $0.216 | $0.647 | 76 | 100.0% | 1.05M |
| Hy4 preview | Hunyuan | - | - | $0.834 | $2.50 | 12 | 97.8% | 1.05M |
| Nemotron 3 Ultra | Nemotron | - | - | $0.50 | $2.20 | 89 | 100.0% | 262K |
| Kimi K3 | Kimi | - | - | $2.85 | $14.25 | 50 | 98.7% | 1.05M |
| GLM 5.2 | GLM | - | - | $0.563 | $1.80 | 91 | 100.0% | 1.05M |
| MiniMax M3 | MiniMax | - | - | $0.28 | $1.10 | 40 | 99.5% | 524K |
| DeepSeek V4 Pro 0813 | DeepSeek | - | - | $1.30 | $2.60 | 64 | 99.7% | 1.05M |
| Nemotron 3.5 Lightning | Nemotron | - | - | $0.06 | $0.16 | 274 | 99.4% | 262K |
| DeepSeek V4 Pro 0423 | DeepSeek | - | - | $1.30 | $2.60 | 33 | 100.0% | 1.05M |
| Nemotron 3 Super | Nemotron | - | - | $0.085 | $0.40 | 59 | 100.0% | 262K |
| Qwen3.8 27B | Qwen | - | - | $0.15 | $1.88 | 58 | 94.7% | 262K |
| Gemma 4 26B A4B | Gemma | - | - | $0.07 | $0.34 | 69 | 99.2% | 262K |
| DeepSeek V3.2 | DeepSeek | - | - | $0.26 | $0.38 | 19 | 100.0% | 164K |
| Inkling | Other open-weight | - | - | $0.95 | $4.05 | 126.5 | 99.3% | 524K |
| gpt-oss-20b | gpt-oss | - | - | $0.03 | $0.14 | 122 | 99.7% | 131K |
| Kimi K2.6 | Kimi | - | - | $0.75 | $3.50 | 13 | 95.9% | 262K |
| Inkling Small | Other open-weight | - | - | $0.45 | $1.20 | 109 | 99.8% | 524K |
| Ling 3.0 Flash | Other open-weight | - | - | $0.06 | $0.18 | 65 | 94.0% | 131K |
| Qwen3 235B A22B Instruct 2507 | Qwen | - | - | $0.09 | $0.55 | 23 | 81.3% | 262K |
| Qwen3.8 2.4T A95B | Qwen | - | - | $2.00 | $6.00 | 138 | 99.8% | 262K |
| Qwen3.6 35B A3B | Qwen | - | - | $0.10 | $0.95 | 8 | 96.9% | 262K |
| Gemma 3 27B | Gemma | - | - | $0.08 | $0.16 | 24 | 100.0% | 131K |
| Ling 3.0 Flash VL | Other open-weight | - | - | $0.06 | $0.18 | 57 | 99.9% | 131K |
| GLM 4.7 | GLM | - | - | $0.40 | $1.75 | 12 | 87.6% | 203K |
| DeepSeek V3.1 | DeepSeek | - | - | $0.25 | $0.95 | 10 | 100.0% | 164K |
| Llama 3.1 8B Instruct | Llama | - | - | $0.02 | $0.04 | 39 | 99.7% | 131K |
| Llama 3.3 70B Instruct | Llama | - | - | $0.10 | $0.32 | 10 | 95.3% | 131K |
| Qwen3.5-9B | Qwen | - | - | $0.10 | $0.15 | 35 | 99.2% | 262K |
| Muse Glimmer 30B | Other open-weight | - | - | $0.30 | $1.20 | 120 | 100.0% | 131K |
| Qwen3.5 397B A17B | Qwen | - | - | $0.45 | $3.00 | 43 | 99.3% | 262K |
| Mistral Small 3.2 24B | Mistral | - | - | $0.075 | $0.20 | 31 | 100.0% | 128K |
| Qwen3 Next 80B A3B Instruct | Qwen | - | - | $0.09 | $1.10 | 56 | 99.1% | 262K |
| Qwen3 VL 235B A22B Instruct | Qwen | - | - | $0.20 | $0.88 | 15 | 97.3% | 262K |
| DeepSeek V3 | DeepSeek | - | - | $0.32 | $0.89 | 25 | 82.6% | 164K |
| GLM 4.6 | GLM | - | - | $0.50 | $2.00 | 19 | 99.5% | 203K |
| Granite 4.2 8B | Other open-weight | - | - | $0.06 | $0.25 | 62 | 100.0% | 131K |
| Qwen3.6 27B | Qwen | - | - | $0.32 | $3.20 | 56 | 99.9% | 262K |
| Qwen3 32B | Qwen | - | - | $0.08 | $0.28 | 26 | 98.9% | 41K |
| Qwen3 Coder 480B A35B | Qwen | - | - | $0.30 | $1.00 | 94 | 93.2% | 262K |
| Mistral Small 3 | Mistral | - | - | $0.05 | $0.08 | 48 | 100.0% | 33K |
| Gemma 3 12B | Gemma | - | - | $0.05 | $0.15 | 34 | 99.6% | 131K |
| Qwen3.5-35B-A3B | Qwen | - | - | $0.14 | $1.00 | 125 | 99.1% | 262K |
| Qwen3.5-27B | Qwen | - | - | $0.26 | $2.60 | 51 | 100.0% | 262K |
| R1 0528 | DeepSeek | - | - | $0.50 | $2.15 | 31 | 99.9% | 164K |
| Llama 4 Scout | Llama | - | - | $0.10 | $0.30 | 35 | 100.0% | 328K |
| Qwen3 VL 30B A3B Instruct | Qwen | - | - | $0.15 | $0.60 | 27 | 99.4% | 262K |
| Qwen3 14B | Qwen | - | - | $0.12 | $0.24 | 23 | 99.6% | 41K |
| Phi 4 | Other open-weight | - | - | $0.07 | $0.14 | 70 | 100.0% | 16K |
| Llama 3 8B Lunaris | Other open-weight | - | - | $0.04 | $0.05 | 90 | 100.0% | 8K |
| Qwen2.5 72B Instruct | Qwen | - | - | $0.36 | $0.40 | 42 | 96.7% | 33K |
| Nemotron 3.5 Content Safety | Nemotron | - | - | $0.20 | $0.20 | 217 | 98.6% | 131K |
| Qwen3 30B A3B | Qwen | - | - | $0.12 | $0.50 | 51 | 100.0% | 41K |
| Hermes 3 70B Instruct | Other open-weight | - | - | $0.70 | $0.70 | 30 | 100.0% | 131K |
| Llama 3.1 70B Instruct | Llama | - | - | $0.40 | $0.40 | 35.5 | 99.1% | 131K |
| Llama Guard 4 12B | Llama | - | - | $0.18 | $0.18 | - | 99.9% | 164K |
| MythoMax 13B | Other open-weight | - | - | $0.40 | $0.40 | 47 | 100.0% | 4K |
| Gemma 3 4B | Gemma | - | - | $0.05 | $0.10 | 24 | 100.0% | 131K |
| Hermes 3 405B Instruct | Other open-weight | - | - | $1.00 | $1.00 | 24 | 100.0% | 131K |
| Llama 3.1 Euryale 70B v2.2 | Other open-weight | - | - | $0.85 | $0.85 | 42 | 99.9% | 131K |
List prices per million tokens as of October 9, 2026. Share: of all tokens served on that model; a dash where the volume is not reported.
Company
- Website
- deepinfra.com
- Headquarters
- Palo Alto, CA
- Compute
- Owns its GPUs and runs them in colocation across U.S. sites; opened a 1.7 MW Toronto site with over 1,000 NVIDIA B300 GPUs in July 2026.
- Latest financing
- $107M Series B co-led by 500 Global and Georges Harik, with NVIDIA and Supermicro participating (May 2026).
- Status page
- status.deepinfra.com
Sources: DCD, StorageReview, VCA.
Finance your GPUs
Serving tokens on your own GPUs? Tell us the cluster, the models you serve and the contracts behind them, and we will share it with partner funders that fit.
Prefer email? hello@amcompute.com