Phi API Pricing Comparison 2026
Microsoft's Phi models are compact, efficient LLMs designed for on-device and edge deployment. Strong performance relative to their small parameter counts.
21
Models
3
Providers
$0.063
Cheapest/1M
0
Open Weight
What is the cheapest Phi model?
The cheapest Phi API is Phi 4 Multimodal Instruct via Deepinfra at $0.063 per million tokens (blended). There are 21 Phi models across 3 providers.
All Phi Models by Price
21 models| # | Model | Input | Output |
|---|---|---|---|
| 9 | Phi-4-reasoning | $0.125 | $0.500 |
| 1 | Phi 4 Multimodal Instruct | $0.050 | $0.100 |
| 5 | Phi-4-mini-instruct | $0.075 | $0.300 |
| 6 | Phi-4-mini-reasoning | $0.080 | $0.320 |
| 2 | Microsoft: Phi 4 | $0.070 | $0.140 |
| 12 | Phi-3.5-mini-instruct | $0.130 | $0.520 |
| 15 | Phi 3.5 MoE Instruct | $0.160 | $0.640 |
| 7 | Phi-3-medium-4k-instruct | $0.140 | $0.140 |
| 16 | Phi-3-medium-128k-instruct | $0.170 | $0.680 |
| 4 | Phi-3-mini-128k-instruct | $0.100 | $0.100 |
| 8 | Phi 3 Vision 128K Instruct | $0.200 | $0.200 |
| 10 | Phi-3.5-vision-instruct | $0.130 | $0.520 |
| 11 | Phi-3-mini-4k-instruct | $0.130 | $0.520 |
| 13 | Phi-3-small-8k-instruct | $0.150 | $0.600 |
| 14 | Phi-3-small-128k-instruct | $0.150 | $0.600 |
| 3 | Phi 2 3B | $0.100 | $0.100 |
| 17 | Dolphin 2.6 Mixtral 8x7b | $0.500 | $0.500 |
| 18 | Dolphin 2.9 2 Qwen2 72B | $0.900 | $0.900 |
| 19 | Phind Code Llama 34B Python V1 | $0.900 | $0.900 |
| 20 | Phind Code Llama 34B V1 | $0.900 | $0.900 |
| 21 | Phind Code Llama 34B V2 | $0.900 | $0.900 |
Frequently Asked Questions
The cheapest Phi model is Phi 4 Multimodal Instruct via Deepinfra at $0.063 per million tokens (blended rate). Prices vary across providers — Sector HQ tracks them all.