Phi 4 Mini

Phi 4 Mini Instruct is a lightweight (3.8B parameters) open model built upon synthetic data and filtered web data, focusing on high-quality reasoning. It supports a 128K token context length and is enhanced for instruction adherence and safety via supervised fine-tuning and direct preference optimization.

GSM8k

88.6%

i
ARC-C

83.7%

i
BoolQ

81.2%

i
OpenBookQA

79.2%

i
PIQA

77.6%

i
Social IQa

72.5%

i
BIG-Bench Hard

70.4%

i
HellaSwag

69.1%

i
MMLU

67.3%

i
Winogrande

67.0%

i
TruthfulQA

66.4%

i
MATH

64.0%

i
MGSM

63.9%

i
MMLU-Pro

52.8%

i
Multilingual MMLU

49.3%

i
Arena Hard

32.8%

i
GPQA

25.2%

i

Pricing, uptime, and speed via OpenRouter — updated Jun 30, 2026, 07:11 PM.

Provider	Status	Input	Output	Limits	Uptime	Speed	Notes
WandB	available	$0.08/Mtok cache $0.08/Mtok	$0.35/Mtok	128K tokens context 128K tokens max output	100.0% 5m 100.0%	181 ms p50 TTFT 47 tok/s p50	bf16