Microsoft model product
phi-4 is a state-of-the-art open model built to excel at advanced reasoning, coding, and knowledge tasks. It leverages a blend of synthetic data, filtered web data, academic texts, and supervised fine-tuning for precision, alignment, and safety.
Updated Aug 10, 2026. Default version: Phi 4
Structured fields from the published default version.
This profile uses the latest version under this unique model that has a calculated LLMBoard score. Arena and price are excluded.
Published benchmark records for the scored version Phi 4.
Preference and agent-evaluation signals from published Arena datasets.
| text | japanese | 198 | 1159.2 | 500 | N/A | |
| text style control | japanese | 205 | 1195.1 | 500 | N/A | |
| text | korean | 217 | 1151.3 | 319 | N/A | |
| text | french | 221 | 1223.4 | 205 | N/A | |
| text | spanish | 221 | 1231.4 | 127 | N/A | |
| text | german | 222 | 1222.8 | 523 | N/A | |
| text style control | korean | 223 | 1196.2 | 319 | N/A | |
| text style control | spanish | 225 | 1265.9 | 127 | N/A | |
| text style control | french | 226 | 1265.3 | 205 | N/A | |
| text style control | german | 228 | 1256.3 | 523 | N/A | |
| text | industry mathematical | 248 | 1253.8 | 2,350 | N/A | |
| text style control | industry mathematical | 249 | 1278.4 | 2,350 | N/A | |
| text | math | 253 | 1245.8 | 2,764 | N/A | |
| text | expert | 256 | 1202.1 | 1,124 | N/A | |
| text style control | expert | 257 | 1266.7 | 1,124 | N/A | |
| text style control | math | 258 | 1264.9 | 2,764 | N/A | |
| text | industry legal and government | 266 | 1241.1 | 1,341 | N/A | |
| text style control | industry legal and government | 266 | 1298.4 | 1,341 | N/A | |
| text | chinese | 267 | 1211.3 | 1,503 | N/A | |
| text | industry medicine and healthcare | 267 | 1201.0 | 1,037 | N/A | |
| text | coding | 268 | 1231.8 | 3,305 | N/A | |
| text | hard prompts | 269 | 1219.8 | 5,747 | N/A | |
| text | hard prompts english | 270 | 1231.3 | 3,804 | N/A | |
| text | russian | 271 | 1207.9 | 2,576 | N/A | |
| text style control | industry medicine and healthcare | 272 | 1272.3 | 1,037 | N/A | |
| text style control | hard prompts english | 273 | 1290.2 | 3,804 | N/A | |
| text | industry software and it services | 274 | 1228.8 | 5,580 | N/A | |
| text style control | chinese | 274 | 1267.0 | 1,503 | N/A | |
| text | industry business and management and financial operations | 275 | 1198.6 | 2,438 | N/A | |
| text | instruction following | 275 | 1201.7 | 9,162 | N/A | |
| text style control | coding | 275 | 1306.4 | 3,305 | N/A | |
| text | longer query | 277 | 1217.5 | 2,896 | N/A | |
| text style control | hard prompts | 278 | 1277.7 | 5,747 | N/A | |
| text style control | russian | 278 | 1245.7 | 2,576 | N/A | |
| text | multi turn | 279 | 1205.9 | 3,517 | N/A | |
| text | non english | 279 | 1196.8 | 10,025 | N/A | |
| text style control | industry software and it services | 279 | 1290.1 | 5,580 | N/A | |
| text | english | 280 | 1231.0 | 14,101 | N/A | |
| text style control | industry business and management and financial operations | 280 | 1256.2 | 2,438 | N/A | |
| text style control | longer query | 281 | 1266.5 | 2,896 | N/A | |
| text | overall | 283 | 1216.9 | 24,126 | N/A | |
| text | exclude ties | 283 | 1100.7 | 15,789 | N/A | |
| text style control | instruction following | 283 | 1245.3 | 9,162 | N/A | |
| text style control | multi turn | 283 | 1241.6 | 3,517 | N/A | |
| text | creative writing | 285 | 1182.0 | 4,062 | N/A | |
| text | industry life and physical and social science | 285 | 1206.8 | 4,245 | N/A | |
| text | industry writing and literature and language | 286 | 1197.9 | 6,470 | N/A | |
| text style control | english | 287 | 1271.8 | 14,101 | N/A | |
| text | industry entertainment and sports and media | 288 | 1162.5 | 4,486 | N/A | |
| text style control | non english | 288 | 1233.0 | 10,025 | N/A | |
| text style control | overall | 289 | 1256.2 | 24,126 | N/A | |
| text style control | exclude ties | 289 | 1156.9 | 15,789 | N/A | |
| text style control | industry writing and literature and language | 289 | 1229.0 | 6,470 | N/A | |
| text style control | creative writing | 290 | 1210.2 | 4,062 | N/A | |
| text style control | industry life and physical and social science | 290 | 1264.6 | 4,245 | N/A | |
| text style control | industry entertainment and sports and media | 291 | 1202.5 | 4,486 | N/A |
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
| OpenRouter | microsoft/phi-4 | global | $0.07 | $0.14 | 16.4K | |
| Azure Cognitive Services | phi-4 | global | $0.125 | $0.50 | 128K | |
| Azure | phi-4 | global | $0.125 | $0.50 | 128K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
All published versions linked to this unique model. The score columns identify the version used by the current overall ranking.
| Phi 4 | 36.9 | 14.7B | 16K | 16K | No | MIT |
A concise description based on the published model registry.
phi-4 is a state-of-the-art open model built to excel at advanced reasoning, coding, and knowledge tasks. It leverages a blend of synthetic data, filtered web data, academic texts, and supervised fine-tuning for precision, alignment, and safety.
Use the benchmark, Arena and pricing sections above as separate evidence. A missing field means the current data snapshot does not support that claim.
Data snapshot: 2026-08-07. Editorial model content is not available in the backend.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest published LLMBoard score.
Common questions about Phi 4.
Phi 4's default version was released on Dec 12, 2024.
Phi 4's official API price is $0.125 per million input tokens and $0.50 per million output tokens via Azure. The lowest tracked third-party offer starts at $0.07 input and $0.14 output via OpenRouter.
Phi 4 is published under Microsoft in the model registry.
The default version has a 16K token context window.
No. The default version is not marked as having publicly available weights.
3 published provider offerings are linked to the default version.
Nearby ranked alternatives include DiffusionGemma 26B A4B, DeepSeek-V3, Qwen2 72B.