Microsoft model product
Phi 4 Reasoning
Phi-4-reasoning is a state-of-the-art open-weight reasoning model finetuned from Phi-4 using supervised fine-tuning on a dataset of chain-of-thought traces and reinforcement learning.
Updated Aug 12, 2026. Default version: Phi 4 Reasoning
LLMBoard score22.0Phi 4 Reasoning
Coverage60%9 benchmark families
Context windowN/ATokens
Official input price$0.125Azure API
Phi 4 Reasoning Specifications
Technical details for the model's default version.
- Version
- Phi 4 Reasoning
- Released
- Apr 30, 2025
- Knowledge cutoff
- Mar 1, 2025
- Parameters
- 14B
- Context window
- N/A
- Max output
- N/A
- Inputs
- text
- Outputs
- text
- Open weights
- No
- License
- MIT
Phi 4 Reasoning Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Phi 4 Reasoning Benchmark Results
Benchmark scores for Phi 4 Reasoning.
| HumanEval+ | 0.9 | 1 | 10 | 100.0% | C | |
| FlenQA | 1.0 | 2 | 2 | 0.0% | C | |
| OmniMath | 0.8 | 2 | 2 | 0.0% | C | |
| PhiBench | 0.7 | 2 | 3 | 50.0% | C | |
| Arena Hard | 0.7 | 10 | 26 | 64.0% | C | |
| AIME 2024 | 0.8 | 34 | 53 | 36.5% | C | |
| LiveCodeBench | 0.5 | 39 | 73 | 47.2% | C | |
| IFEval | 0.8 | 45 | 65 | 31.3% | C | |
| MMLU-Pro | 0.7 | 74 | 129 | 43.0% | C | |
| AIME 2025 | 0.6 | 95 | 114 | 16.8% | C | |
| GPQA | 0.7 | 146 | 234 | 37.8% | C |
Phi 4 Reasoning Arena Results
Preference and agent-evaluation results for the default version.
No Arena results
The default version does not have a matching Arena result yet.
Phi 4 Reasoning Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- $0.125 input, $0.50 output per 1M
- Official provider
- Azure
- Lowest third-party
- From $0.125 input, $0.50 output per 1M via Azure Cognitive Services
- Tracked offerings
- 2
| Azure Cognitive Services | phi-4-reasoning | global | $0.125 | $0.50 | 32K | |
| Azure | phi-4-reasoning | global | $0.125 | $0.50 | 32K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Phi 4 Reasoning Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| Phi 4 Reasoning | 22.0 | 14B | N/A | N/A | No | MIT |
What is Phi 4 Reasoning?
Key information about Phi 4 Reasoning and its available data.
Phi-4-reasoning is a state-of-the-art open-weight reasoning model finetuned from Phi-4 using supervised fine-tuning on a dataset of chain-of-thought traces and reinforcement learning. It focuses on math, science, and coding skills.
Data as of 2026-08-11.
Phi 4 Reasoning vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to Phi 4 Reasoning
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
MI
Phi 4 Reasoning Plus
Microsoft
MI
Phi 4 Mini Reasoning
Microsoft
MI
Phi 4
Microsoft
MI
Phi 4 multimodal
Microsoft
MI
Phi 3.5 MoE
Microsoft
MI
Phi 3.5 vision
Microsoft
FAQ
Common questions about Phi 4 Reasoning.
When was Phi 4 Reasoning released?
Phi 4 Reasoning's default version was released on Apr 30, 2025.
How much does Phi 4 Reasoning cost?
Phi 4 Reasoning'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 $0.125 input and $0.50 output via Azure Cognitive Services.
Who created Phi 4 Reasoning?
Phi 4 Reasoning was created by Microsoft.
What is the context window for Phi 4 Reasoning?
A context window is not available for the default version.
Is Phi 4 Reasoning open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer Phi 4 Reasoning?
2 provider offerings are linked to the default version.
What models should I compare Phi 4 Reasoning with?
Nearby ranked alternatives include Qwen2.5 VL 72B, Llama 3.3 Nemotron Super 49B, MiniCPM SALA.