MicrosoftMicrosoft

Phi 3 Mini 3.8B

現行
250.9Kダウンロード1.7KいいねApr 2024公開日128K トークンコンテキストMITライセンス42 基本品質

Phi 3 Mini 3.8B (3.799999952316284B parameters) requires approximately 10.0 GB of VRAM with Q4_K_M quantization. For the best balance of quality and speed, we recommend hardware with at least 12 GB of VRAM.

はじめに

— コピー&ペーストでローカル実行

Copy-paste commands to run Phi 3 Mini 3.8B on your machine.

Run

ollama run phi3:mini

Quick specs

Parameters3.8B
Architecturedense
Context128K tokens
Modalitytext
Min RAM1.5 GB
Rec. RAM2.3 GB (Q4_K_M)
LicenseMIT
FamilyPhi
Chat

About this model

The Phi-3-Mini-128K-Instruct is a 3.8 billion-parameter, lightweight, state-of-the-art open model trained using the Phi-3 datasets. This dataset includes both synthetic data and filtered publicly available website data, with an emphasis on high-quality and reasoning-dense properties. The model belongs to the Phi-3 family with the Mini version in two variants 4K and 128K which is the context length (in tokens) that it can support.

  • Memory/compute constrained environments
  • Latency bound scenarios
  • Strong reasoning (especially code, math and logic)

関連モデル

あなたのハードウェア

検出中...

おすすめ

最適なハードウェア

Phi 3 Mini 3.8Bのおすすめ

このモデルを実行

量子化オプション

量子化レベル別VRAM推定値

No hardware detected — fit column shows raw VRAM estimates

QuantBitsVRAMQualityFit
Q2_K
2
1.5 GB
Low
Q3_K_S
3
1.9 GB
Low
NVFP4
4
2.1 GB
Medium
Q4_K_M
4
2.3 GB
Medium
Q5_K_M
5
2.7 GB
High
Q6_K
6
3.1 GB
High
Q8_0
8
4.1 GB
Very High
F16
16
7.8 GB
Maximum

Quality benchmarks

Phi 3 Mini 3.8B benchmark scores

Benchmark verified

Coding

SWE-bench Verified
HumanEval+60.4%
Aider Polyglot
LiveCodeBench

Reasoning

MMLU-Pro69.7%
GPQA Diamond29.7%
MATH-50014.0%
ARC Challenge85.5%

General

Chatbot Arena
IFEval59.8%

Source: official · 2024-04-22

ハードウェア互換性

全ハードウェアの適合度推定

カリキュレーターを開く

Computing compatibility...

メモリ内訳

Reference: RTX 2060 6GB

Weights2.3 GB
KV Cache5.9 GB
Runtime1.2 GB
Headroom0.6 GB

よくある質問

FAQ — Phi 3 Mini 3.8B

関連項目