Phi-3-mini-128k-instruct
Microsoft- Arena score
- 1050
- Among open models
- #144of 148
- Among all on Arena
- #399
- Votes
- 20,685
Measured on Phi-3 Mini 128K Instruct, the served version, not a local build. LMArena, CC BY 4.0, published Nov 11, 2026.
Score by category
CategoryScoreOpen rankArena rankVotes
- Math1089#129#3632.8K
- English1083#143#39712.6K
- Applied math1083#132#3662.6K
- Spanish1058#106#301366
- Hard prompts, English1051#141#3904.1K
- Writing & literature1040#140#3954.6K
- Coding1039#142#3913.9K
- French1039#102#294372
- Software & IT1034#144#3986.1K
- Hard prompts1028#142#3926.2K
- Creative writing1024#142#3992.9K
- Science1023#141#3953.4K
- Instructions1023#142#3977.4K
- Chinese1016#133#3722K
- Media & sports1009#141#3972.9K
- Legal & government1006#142#3831K
- German1006#107#313727
- Russian1004#136#3751.8K
- Non-English1000#141#3918.1K
- Medicine & health996#138#3791.1K
- Long prompts996#143#3952.1K
- Business & finance990#145#3952.1K
- Multi-turn989#143#4012.5K
- Expert984#137#3681.1K
- Japanese899#92#286385
- Korean856#98#289260
- Excluding ties845#143#39813.8K
| Model | Params | Score |
|---|---|---|
| 3.8B | 1050 | |
| 3.8B | 1080 |
About as strong, smaller
Compare
Llama-3.2-1B-InstructMeta Llama10551.2B
Built on it
Ways to run it
- GGUF · llama.cpp, Ollama, LM Studiobartowski/Phi-3.1-mini-128k-instruct-GGUF51 builds
- MLX · Apple silicon: mlx-lm, LM Studiodeimagjas/Phi-3.5-mini-instruct-4bit-sft1 build
- AWQ · vLLM, SGLangsolidrust/Phi-3-mini-128k-instruct-AWQ2 builds
- GPTQ · vLLM, SGLangjustinthelaw/Phi-3-mini-128k-instruct-4bit-128g-GPTQ1 build
Most downloaded
- Phi-3.1-mini-128k-instruct-GGUFbartowski · Quantization6.6K downloads
- Phi-3-mini-128k-instruct-GGUFsecond-state · Quantization3.6K downloads
- Phi-3-mini-128k-instruct-IMat-GGUFlegraphista · Quantization2.6K downloads
- Phi-3-mini-128k-instruct-iMat-GGUFqwp4w3hyb · Quantization1.4K downloads
- phi3calcuis · Quantization1K downloads
BenchClub runs
Not runnable yet
Runs open once a Phi-3-mini-128k-instruct build is registered.
More from Microsoft
See more
Phi-3.5-mini-instruct324.4K3.8B

Phi-3-mini-4k-instruct314.8K3.8B

phi-4205.7K14.7B

phi-2177.4K2.8B
Other 3–9B models
See more
NVIDIA-Nemotron-3-Nano-4B-BF16NVIDIA1.9M4.0B
granite-4.1-8bIBM Granite79.4K8.8B
Molmo2-8BAi247.6K8.7B
Spark-X2.6-4BXHToken32.9K4.1B