Provider Comparison · 28 Shared Models
Deepinfra vs Hugging Face
Green highlights mark broader sourced coverage or lower directly comparable prices. They do not imply independently measured performance.
Routing Surface
Side-by-Side Provider Facts
| Field | Deepinfra | Hugging Face |
|---|---|---|
| Routing mode | direct | gateway |
| API protocol | openai | openai |
| API base URL | https://api.deepinfra.com/v1/openai | https://router.huggingface.co/v1 |
| Route-ready models | 49 | 41 |
| Tracked model offerings | 49 | 104 |
| Lowest input / 1M | $0.010 | Unknown |
| Lowest output / 1M | $0.10 | Unknown |
| Service tiers | Standard | Standard |
| Regions | Global | Global, Us |
7 comparable fields · 5 material differences · Interactive only until provider-comparison indexing and telemetry gates are defined
Deepinfra Route Contract
https://api.deepinfra.com/v1/openaiHugging Face Route Contract
https://router.huggingface.co/v1Common Coverage
Models Offered by Both Providers
| Model | Canonical ID | Coverage |
|---|---|---|
| MiniMax M2.7 | MiniMaxAI/MiniMax-M2.7 | Available from both provider routes |
| MiniMax M3 | MiniMaxAI/MiniMax-M3 | Available from both provider routes |
| Qwen3.5 35B A3B | Qwen/Qwen3.5-35B-A3B | Available from both provider routes |
| Qwen3.5 397B A17B | Qwen/Qwen3.5-397B-A17B | Available from both provider routes |
| Qwen3.8 2.4T A95B | Qwen/Qwen3.8-2.4T-A95B | Available from both provider routes |
| Qwen3.8 27B | Qwen/Qwen3.8-27B | Available from both provider routes |
| DeepSeek V3 | deepseek-ai/DeepSeek-V3 | Available from both provider routes |
| DeepSeek V3.2 | deepseek-ai/DeepSeek-V3.2 | Available from both provider routes |
| DeepSeek V4 Flash | deepseek-ai/DeepSeek-V4-Flash | Available from both provider routes |
| DeepSeek V4 Flash Vision Exp | deepseek-ai/DeepSeek-V4-Flash-Vision-Exp | Available from both provider routes |
| DeepSeek V4 Pro | deepseek-ai/DeepSeek-V4-Pro | Available from both provider routes |
| Granite 4.2 30B | ibm-granite/granite-4.2-30b | Available from both provider routes |
| Granite 4.2 3B | ibm-granite/granite-4.2-3b | Available from both provider routes |
| Granite 4.2 8B | ibm-granite/granite-4.2-8b | Available from both provider routes |
| Llama 4 Scout 17B 16E Instruct | meta-llama/Llama-4-Scout-17B-16E-Instruct | Available from both provider routes |
| phi-4 | microsoft/phi-4 | Available from both provider routes |
| Kimi K2.5 | moonshotai/Kimi-K2.5 | Available from both provider routes |
| Kimi K2.6 | moonshotai/Kimi-K2.6 | Available from both provider routes |
| Kimi K3 | moonshotai/Kimi-K3 | Available from both provider routes |
| gpt-oss-120b | openai/gpt-oss-120b | Available from both provider routes |
| gpt-oss-20b | openai/gpt-oss-20b | Available from both provider routes |
| Hy3 | tencent/Hy3 | Available from both provider routes |
| Hy4 preview | tencent/Hy4-preview | Available from both provider routes |
| GLM 5 | zai-org/GLM-5 | Available from both provider routes |
| GLM 5.1 | zai-org/GLM-5.1 | Available from both provider routes |
| GLM 5.2 | zai-org/GLM-5.2 | Available from both provider routes |
| GLM 5.3 | zai-org/glm-5.3 | Available from both provider routes |
| GLM 5.3 Flash | zai-org/glm-5.3-flash | Available from both provider routes |
Provider Evidence
Sources and Freshness
Questions
Deepinfra vs Hugging Face FAQs
Is Deepinfra or Hugging Face cheaper for inference?+
The lowest sourced input prices are $0.010 for Deepinfra and — for Hugging Face. Lowest output prices are $0.10 and —. These minima may refer to different models or tiers.
Which offers more route-ready models, Deepinfra or Hugging Face?+
Deepinfra has 49 route-ready models and Hugging Face has 41, so Deepinfra has broader current route-ready coverage.
Do Deepinfra and Hugging Face offer the same models?+
They currently share 28 tracked models, including MiniMax M2.7, MiniMax M3, Qwen3.5 35B A3B, Qwen3.5 397B A17B, and Qwen3.8 2.4T A95B.
Can I use the same API protocol with Deepinfra and Hugging Face?+
Both are currently documented with the openai API protocol, although base URLs, model IDs, and authentication still differ.
Should I choose Deepinfra or Hugging Face?+
Choose based on the exact model route, price observation, protocol, regions, and service tier you need. The green highlights identify lower sourced minima or broader documented coverage; they do not measure latency, reliability, or output quality.
Which lists more AI models, Deepinfra or Hugging Face?+
Deepinfra lists 49 tracked models; Hugging Face lists 104, so Hugging Face has the larger catalog. Catalog size is separate from route readiness.
Is Deepinfra or Hugging Face a gateway?+
Deepinfra is recorded as a direct route; Hugging Face is recorded as a gateway. Gateways aggregate model routes, while direct providers serve their own endpoint surface.
Can I switch between Deepinfra and Hugging Face without changing model IDs?+
Do not assume model IDs are portable. Even with 28 shared tracked models, base URLs, authentication, provider-specific identifiers, tiers, and request features can differ.
Are Deepinfra and Hugging Face prices directly comparable?+
Only at the exact model, tier, region, and token type. The headline minima above can come from different models or service tiers, so they are useful discovery signals rather than an apples-to-apples bill estimate.
How current is the Deepinfra vs Hugging Face comparison?+
Deepinfra was last observed Oct 19, 2026 and Hugging Face Sep 2, 2026. The source register below identifies the records behind the comparison.