Model
Model explorer

Phi-4 (14B)

OPEN
Microsoft · Phi-4 family · released Dec 12, 2024

Flagship 14B trained on ~9.8T tokens of heavy synthetic data; surpassed teacher GPT-4o on GPQA and MATH; weights fully opened on HF Jan 8, 2025.

ReasoningCodingVisionFunction callingTool useAgentic
1016.7
Elo · rank #209
Parameters
14B
Active params
14B (dense)
Context
16K tokens
Architecture
Dense decoder-only Transformer (phi-3-medium-style architecture, tiktoken tokenizer)
License
MIT
Languages
—
API price (in/out)
$0.125 / $0.5
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
Arena-HardHuman preference75.4%#30
best: Qwen3-235B-A22B (Non-Thinking) · 96.1%
DROPReasoning75.5#36
best: Hunyuan-T1 · 93.1
GPQA DiamondReasoning56.1%#173
best: GPT-5.6 · 94.6%
HumanEval+Coding82.8%#12
best: Mistral Small 3.2 (24B) · 92.9%
HumanEvalCoding82.6%#48
best: Claude Opus 4.5 · 99.4%
IFEvalReasoning63.0%#118
best: Gemma 4 26B A4B · 98.5%
MATH-500Math80.4%#68
best: GPT-5 · 99.4%
MGSMMath80.6%#27
best: OpenAI o4-mini · 93.7%
MMLU-ProKnowledge70.4%#91
best: Claude Fable 5 · 91.5%
MMLUKnowledge84.8%#61
best: OpenAI o3 · 92.9%
SimpleQAKnowledge3.0%#47
best: GPT-4.5 · 62.5%
Run it locally
VRAM @ Q4
11 GB
VRAM @ FP16
29 GB
Fits on (Q4)
RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Throughput data unavailable.
Quantizations
GGUF Q4 · MLX 4-bit
Phi-4 family
Elo progression across releases
Phi-4 (14B)Dec 20241016.7Phi-4-mini (3.8B)Feb 2025529.8Phi-4-mini-instructFeb 2025523.2Phi-4-multimodal (5.6B)Feb 2025842.4Phi-4-multimodal-instructFeb 2025842.4Phi-4-reasoningApr 20251292.1Phi-4-reasoning (14B)Apr 20251292.1
Resources
HFmicrosoft/phi-4GHmicrosoft/PhiCookBookDOCSarxiv.org/abs/2412.08905APIAvailable via Microsoft API
API price $0.125/$0.5 · each benchmark row carries its own source badge (see methodology)