whatmodel.sh / MiniCPM5 2B

MiniCPM5 2B

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

MiniCPM5 2B is an open-weights language model from OpenBMB, released September 2026. Its published size is 2.52B parameters and a 131k-token context window. Licence: apache-2.0.

Catalogue identity notes

OpenBMB's MiniCPM5-2B (openbmb/MiniCPM5-2B @f974000, created 2026-09-06, apache-2.0), the second model of the MiniCPM5 series (MiniCPM5-1B came first). Deep-thinking model (SFT + RL + on-policy distillation).

Model information sources

Release date ↗ · Weights licence ↗

HF safetensors metadata 2,516,756,480 parameters.

63.5±1.1#162 of 719 LLMsfrom 20 benchmark results across 1 source

$ how it compares

The LLMs ranked either side of MiniCPM5 2B, out of 828 with a score, then more from OpenBMB.

#modelscorewhereRAMreleased
184Nova 2 Pro (Preview) Amazon64±2api $3.78/M2025-11
185Grok 3 Mini Beta SpaceXAI (xAI)64±1comparison only2025-04
186MiniCPM5 2B OpenBMB64±1local~3GB2026-09
187Grok 3 Beta SpaceXAI (xAI)63±1comparison only2025-04
188Qwen3-Next 80B-A3B Thinking Alibaba63±1api $0.41/M +local~27GB2025-09
more from OpenBMB
467MiniCPM3-4B OpenBMB48±2local~4GB2024-09
574MiniCPM-V 2.6 OpenBMB43±2local~6.4GB2024-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #127 of 300 · 21% of its score · 5 tests

Beats o3, Gemini 2.5 Pro, Gemini 3.1 Flash-Lite here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Large here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Verified lab46.4% ✎▄··▄▄█··▄·▄█ beats 8/8very hard · avg 12%▇▇ 5.3%sharp · at its level · noisy
BrowseComp-ZH lab43.5% ✎▄▄·▄▄▄·█···█ beats 8/8very hard · avg 13%▇▇ 5.1%at its level
SWE-bench Pro (split unspecified) lab14.4% ✎█▂···▂···▄·▂ beats 7/8very hard · avg 4.3%▇▇ 5.0%sharp · noisy
BFCL V4 (December 2025 archive) lab66.6% ✎▄·█·▄▄▄▄·▄·▄ beats 8/8medium · avg 43%▇ 3.7%at its level
τ²-bench Telecom lab97.1% ✎▁·▅█▆▅▂▃▃▃▃▁ beats 79/89hard · avg 32%▇ 1.9%sharp · noisy · small run
coding▁▁▁▃▆██▇▆▄▃▁ #154 of 458 · 34% of its score · 5 tests

Beats Gemini 2.5 Pro, Gemini 3.1 Flash-Lite, Qwen3-Coder-Next (80B-A3B) here, though they rank higher overall. Trails Ministral 3 8B Reasoning 2512, Phi-4-reasoning 14B, DeepSeek V3 (Dec '24) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LiveCodeBench Pro 2025Q2 (Medium) lab17.5% ✎█···▂··▄▂··▃ beats 8/8very hard · avg 2.0%▇▇▇▇ 11%sharp · noisy
LiveCodeBench v6 (window unspecified) lab69.1% ✎▃····█·▃▅▃·▃ beats 8/8medium · avg 39%▇▇▇ 6.5%at its level
LiveCodeBench Pro 2025Q2 (Easy) lab68.0% ✎▄···█▄··▄█▄▄ beats 8/8hard · avg 33%▇▇ 5.9%at its level
SWE-bench Verified lab46.4% ✎▄··▄▄█··▄·▄█ beats 8/8very hard · avg 12%▇▇ 5.3%sharp · at its level · noisy
SWE-bench Pro (split unspecified) lab14.4% ✎█▂···▂···▄·▂ beats 7/8very hard · avg 4.3%▇▇ 5.0%sharp · noisy
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #172 of 393 * · 0.2% of its score · 1 test

Trails LFM2.5 1.2B Instruct, LFM2.5 1.2B Thinking, Gemma 3 12B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
IFEval (aggregation unspecified) lab86.7% ✎▄▄····▄··███ beats 3/8easy · avg 77%▇ 0.2%noisy
math▁▂▃▅▆▅▄█▄▁▁▁ #138 of 398 · 29% of its score · 4 tests

Beats Claude Sonnet 4.6, MiniMax-M3, MiMo-V2.5 (310B-A15B) here, though they rank higher overall. Trails Olmo 3 32B Think, Olmo 3.1 32B Think, gpt-oss-20b here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
HMMT February 2026 lab63.8% ✎▅··▃▃·▃··▃·█ beats 7/8hard · avg 30%▇▇▇▇▇ 12%sharp · at its level
AIME 2025 lab86.5% ✎█▄▄▄▄····█·▄ beats 8/8medium · avg 42%▇▇▇ 7.2%sharp
AIME 2026 lab86.5% ✎▄·▄█·▄▄···█▄ beats 8/8medium · avg 48%▇▇ 6.1%sharp
MATH-500 lab94.6% ✎██▄▄▄··▄···▄ beats 6/8very easy · avg 89%▇ 2.7%maxed out
web search█·█▄██·▄▄·▄▄ #3 of 13 * · 5.1% of its score · 1 test

Trails Spark-X2.5 1.7B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BrowseComp-ZH lab43.5% ✎▄▄·▄▄▄·█···█ beats 8/8very hard · avg 13%▇▇ 5.1%at its level
other tests7 not in a task · 26% of its score
testresultvs the others in its rundifficultycountswhy
OJBench lab32.5% ✎▃····▃▅··█▃▃ beats 8/8very hard · avg 12%▇▇ 6.0%at its level
LongBench v2 lab43.7% ✎▃······▅▅▃·█ beats 7/8very hard · avg 32%▇▇ 5.3%at its level
MMLU-Pro lab70.8% ✎▂··▂█·▂▃···▂ beats 7/8medium · avg 59%▇▇ 4.3%typical
SuperGPQA lab40.8% ✎▃·▃·▃▅█····▃ beats 7/8hard · avg 32%▇▇ 3.8%at its level
MMLU-Redux lab84.7% ✎▂···█··▂▃··▂ beats 7/8easy · avg 77%▇ 3.4%typical
IFBench (aggregation unspecified) lab66.3% ✎▅····▃▃·█·▃▃ beats 7/8medium · avg 43%▇ 2.2%at its level
Multi-IF (aggregation unspecified) lab71.8% ✎▄▄···▄·▄·▄██ beats 5/8medium · avg 60%▇ 1.4%blunt

$ where to run it

llama.cpp · recommended

version

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run MiniCPM5 2B · needs ~1.56 GB · from huggingface.co/openbmb (official)
    chat in the terminal
    llama cli -hf openbmb/MiniCPM5-2B-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf openbmb/MiniCPM5-2B-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run MiniCPM5 2B · needs ~1.56 GB · from huggingface.co/openbmb (official)

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/openbmb/MiniCPM5-2B-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run MiniCPM5 2B · needs ~1.56 GB · from huggingface.co/openbmb (official)
    chat in the terminal
    ollama run hf.co/openbmb/MiniCPM5-2B-GGUF:Q4_K_M

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run MiniCPM5 2B · needs ~1.4 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/MiniCPM5-2B-mlx-4Bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/MiniCPM5-2B-mlx-4Bit

Official weights and documentation: huggingface.co/openbmb ↗

$ how the score was worked out · openbmb/MiniCPM5-2B model card 100%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
openbmb/MiniCPM5-2B model card · lab-reported20▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 100%Apache-2.0
prior · pull toward the middle▇ 0%
weighted average= 63.5

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json