Phi-4 Models by Microsoft Research
9 models2024–2026Up to 128k ctxFrom $0.05/1M input
About
Phi-4 is a family of 9 AI models by Microsoft Research, released between 2024 and 2026.
Current Variants
Use-when guidance is derived from seed capabilities, context, release, and replacement fields.
9 in view
Phi-4 Mini ReasoningCurrent
Use when the workload needs reasoning, 128k context, and 3.8B parameters.
2026-05reasoning128k context3.8B parameters
Phi-4 Reasoning Vision 15BCurrent
Use when the workload needs 15B parameters and multimodal inputs.
2026-0315B parametersmultimodal inputs
Phi-4 Mini Flash ReasoningCurrent
Use when the workload needs reasoning, 128k context, and 3.8B parameters.
2025-12reasoning128k context3.8B parameters
Phi 4 Multimodal InstructCurrent
Use when the workload needs multimodal, 128k context, and 5.6B parameters.
2025-01multimodal128k context5.6B parameters
Phi 4 Reasoning PlusCurrent
Use when the workload needs reasoning, 128k context, and 14B parameters.
2025-01reasoning128k context14B parameters
Phi 4 ReasoningCurrent
Use when the workload needs reasoning, 128k context, and 14B parameters.
2025-01reasoning128k context14B parameters
Phi-4 MultimodalCurrent
Use when the workload needs 128k context, 5.6B parameters, and multimodal inputs.
2025-01128k context5.6B parametersmultimodal inputs
Phi-4 14BCurrent
Use when the workload needs 16k context, 14B parameters, and structured outputs.
2024-1216k context14B parametersstructured outputs
Phi-4 MiniCurrent
Use when the workload needs 128k context and 3.8B parameters.
2024-12128k context3.8B parameters
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Phi-4 Mini Reasoning | Use when the workload needs reasoning, 128k context, and 3.8B parameters. | 2026-05 | reasoning128k context3.8B parameters | Current |
| Phi-4 Reasoning Vision 15B | Use when the workload needs 15B parameters and multimodal inputs. | 2026-03 | 15B parametersmultimodal inputs | Current |
| Phi-4 Mini Flash Reasoning | Use when the workload needs reasoning, 128k context, and 3.8B parameters. | 2025-12 | reasoning128k context3.8B parameters | Current |
| Phi 4 Multimodal Instruct | Use when the workload needs multimodal, 128k context, and 5.6B parameters. | 2025-01 | multimodal128k context5.6B parameters | Current |
| Phi 4 Reasoning Plus | Use when the workload needs reasoning, 128k context, and 14B parameters. | 2025-01 | reasoning128k context14B parameters | Current |
| Phi 4 Reasoning | Use when the workload needs reasoning, 128k context, and 14B parameters. | 2025-01 | reasoning128k context14B parameters | Current |
| Phi-4 Multimodal | Use when the workload needs 128k context, 5.6B parameters, and multimodal inputs. | 2025-01 | 128k context5.6B parametersmultimodal inputs | Current |
| Phi-4 14B | Use when the workload needs 16k context, 14B parameters, and structured outputs. | 2024-12 | 16k context14B parametersstructured outputs | Current |
| Phi-4 Mini | Use when the workload needs 128k context and 3.8B parameters. | 2024-12 | 128k context3.8B parameters | Current |
Release Timeline
5 release groups2026-05
1 current
Phi-4 Mini Reasoning
Currentreasoning128k context3.8B parameters
2026-03
1 current
Phi-4 Reasoning Vision 15B
Current15B parametersmultimodal inputs
2025-12
1 current
Phi-4 Mini Flash Reasoning
Currentreasoning128k context3.8B parameters
2025-01
4 current
Phi 4 Multimodal Instruct
Currentmultimodal128k context5.6B parameters
Phi 4 Reasoning
Currentreasoning128k context14B parameters
Phi 4 Reasoning Plus
Currentreasoning128k context14B parameters
Phi-4 Multimodal
Current128k context5.6B parametersmultimodal inputs
2024-12
2 current
Phi-4 14B
Current16k context14B parametersstructured outputs
Phi-4 Mini
Current128k context3.8B parameters
Specifications(9 models)
| Model | Released | Context | Parameters | Vision | Multimodal | Reasoning | Structured Outputs |
|---|---|---|---|---|---|---|---|
| Phi-4 Mini Reasoning | 2026-05 | 128k | 3.8B | No | No | Yes | No |
| Phi-4 Reasoning Vision 15B | 2026-03 | — | 15B | No | Yes | No | No |
| Phi-4 Mini Flash Reasoning | 2025-12 | 128k | 3.8B | No | No | Yes | No |
| Phi 4 Multimodal Instruct | 2025-01 | 128k | 5.6B | Yes | Yes | No | No |
| Phi 4 Reasoning Plus | 2025-01 | 128k | 14B | No | No | Yes | No |
| Phi 4 Reasoning | 2025-01 | 128k | 14B | No | No | Yes | No |
| Phi-4 Multimodal | 2025-01 | 128k | 5.6B | No | Yes | No | No |
| Phi-4 14B | 2024-12 | 16k | 14B | No | No | No | Yes |
| Phi-4 Mini | 2024-12 | 128k | 3.8B | No | No | No | No |
Available From(5 providers)
Pricing
| Model | Provider | Input / 1M | Output / 1M | Type |
|---|---|---|---|---|
| Phi-4 Mini | Novita AI | $0.05 | $0.15 | Serverless |
| Phi-4 14B | OpenRouter | $0.065 | $0.14 | Serverless |
| Phi 4 Reasoning Plus | Microsoft Foundry | $0.125 | $0.5 | Serverless |
| Phi 4 Reasoning Plus | Fireworks AI | $0.5 | $0.5 | Serverless |
| Phi 4 Reasoning | Fireworks AI | $0.5 | $0.5 | Serverless |
| Phi-4 14B | Fireworks AI | $0.9 | $0.9 | Serverless |
| Phi-4 Mini | Fireworks AI | $0.9 | $0.9 | Serverless |
| Phi 4 Multimodal Instruct | Fireworks AI | $0.9 | $0.9 | Serverless |
Frequently Asked Questions
- What is Phi-4 used for?
- Phi-4 is used for reasoning, multimodal, and vision and multimodal work. The family description and listed model capabilities point to those workloads as the best fit.
- How does Phi-4 compare to Harrier?
- Phi-4 by Microsoft Research is strongest where you need reasoning, while Harrier by Microsoft Research is the closest related family to check for embedding. Phi-4 has 9 listed variants and reaches up to 128k context, while Harrier reaches up to 33k context, so compare the specs and pricing tables before choosing a production model.
- Which Phi-4 model should I use?
- For the lowest listed input price, start with Phi-4 Mini through Novita AI at $0.05/1M input tokens. For the most capable/latest local choice, evaluate Phi 4 Multimodal Instruct with 128k context and multimodal inputs.
Related Model Families
Models(9)
Phi-4 Mini Reasoning
2026-05128k3.8B
Reasoning
Phi-4 Reasoning Vision 15B
2026-0315B
MultimodalOpen Source
Phi-4 Mini Flash Reasoning
2025-12128k3.8B1 provider
Reasoning
Phi 4 Multimodal Instruct
2025-01128k5.6B3 providers
Multimodal
Phi 4 Reasoning Plus
2025-01128k14B2 providers
ReasoningOpen Source
Phi 4 Reasoning
2025-01128k14B2 providers
ReasoningOpen Source
Phi-4 Multimodal
2025-01128k5.6B
Multimodal
Phi-4 14B
2024-1216k14B3 providers
Open Source
Phi-4 Mini
2024-12128k3.8B3 providers
Open Source