Microsoft: Phi 4 API pricing
$0.07 per 1M input tokens and $0.14 per 1M output tokens on OpenRouter. Context window: 16,384 tokens.
$0.07input / 1M tokens
$0.14output / 1M tokens
16Kcontext
text->textmodality
What it costs in practice
| Workload | Cost |
|---|---|
| Chatbot: 10,000 messages (1,500 in / 400 out tokens each) | $1.61 |
| RAG: 1,000 queries (8,000 in / 600 out each) | $0.644 |
| Summarize 100 long documents (50,000 in / 1,000 out each) | $0.364 |
Cross-check against the provider's list price
No first-party entry for this model in LiteLLM's price map, so this price comes from OpenRouter only.
Price history
No price change recorded since we started tracking on Nov 17, 2026. We check every 6 hours and log every change.
Run it yourself instead
The open weights are on Hugging Face. See the VRAM phi 4 needs and which GPUs fit it →
Cheaper alternatives with similar context
Text models with at least 16K context and a lower blended price (75% input, 25% output). Price only, not a quality ranking.
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| IBM: Granite 4.2 8B | $0.06 | $0.15 | 128K |
| Qwen: Qwen3.5-9B | $0.073 | $0.107 | 256K |
| Google: Gemma 3 12B | $0.05 | $0.15 | 128K |
| DeepSeek: DeepSeek V4 Flash 0731 (batch) | $0.048 | $0.144 | 1M |
| Meta: Llama 3.2 1B Instruct | $0.027 | $0.201 | 59K |
| OpenAI: GPT-5 Nano (batch) | $0.025 | $0.20 | 391K |
| inclusionAI: Ling 3.1 Flash | $0.045 | $0.135 | 977K |
| Cohere: Command R7B (12-2024) | $0.0375 | $0.15 | 125K |
Source: OpenRouter listing via the public models API. Prices checked Nov 17, 2026. Model ID microsoft/phi-4, listed Jan 10, 2025. Report an error.