Model reference · Synced 2025-04-29
Phi 4 Mini
Phi 4 Mini is an AI model from Nvidia. 131K context window. Capabilities: reasoning, tool calling, multimodal vision, audio, open weights. Available on 6 providers. Cheapest listing: $0 input / $0 output per 1M tokens.
Quick facts
- Cheapest input: $0 per 1M tokens (Nvidia)
- Cheapest output: $0 per 1M tokens
- Context window: 131K tokens
- Max output: 8K tokens
- Release date: 2024-12-01
- Knowledge cutoff: 2024-12
- Capabilities: reasoning, tool calling, multimodal vision, audio, open weights
- Provider count: 6
Provider pricing
Same model, different providers, different prices. Cheapest first.
| Provider | Input / 1M | Output / 1M | Context | Listed |
|---|---|---|---|---|
| Nvidia | $0 | $0 | 131K | 2024-12-01 |
| GitHub Models | $0 | $0 | 128K | 2024-12-11 |
| Azure Cognitive Services | $0.075 | $0.3 | 128K | 2024-12-11 |
| Azure | $0.075 | $0.3 | 128K | 2024-12-11 |
| Weights & Biases | $0.08 | $0.35 | 128K | 2024-12-11 |
| NanoGPT | $0.17 | $0.68 | 128K | 2025-07-26 |
Prices synced daily from models.dev + provider docs.
How to use this model
If you're picking Phi 4 Mini for a project, the three things that matter most:
- Compare it side-by-side with one or two alternatives in the live comparison tool. Pricing differences matter more than benchmarks at scale.
- Pick the cheapest provider that meets your latency / SLA need. Big spread across providers for the same weights.
- Re-evaluate every 3 months. Frontier prices drop fast; a model that's cheapest today may not be in a quarter.
Related models
FAQ
How much does Phi 4 Mini cost? $0 input / $0 output per 1M tokens at the cheapest listing. See the table above for other providers.
What is the context window? 131K tokens.
Which providers offer it? NanoGPT, Weights & Biases, Azure Cognitive Services, Nvidia, Azure, GitHub Models.
Where do these numbers come from? models.dev + provider documentation, synced daily. About the data.