Inference
EN
OpenBMB

MiniCPM5-2B

MiniCPM5-2B is a reasoning model from OpenBMB in the MiniCPM5 family. 21 benchmarks count toward its index, in 6 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2026-09
RankedEnough results for an index.
IndexAll categories, out of 100.37.1±5.0Median 51.0
CoverageIndex weight that has results.85%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.—Median $0.60
Output / 1MUS dollars per 1M output tokens.—Median $2.55
ContextMaximum tokens in one request.131KMedian 262K
EloLMArena rating and rank.—Median 1422

Run locally

2.52B parameters
16-bit
6.3 GB
8-bit
3.8 GB
4-bit
2.6 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 8 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.3 counted38.2181 / 218
CodingCode writing and repair.4 counted32.0195 / 209
ReasoningLogic problems and puzzles.3 counted39.1169 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.5 counted34.7226 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.2 counted44.439 / 63
MathMath problems.4 counted43.9110 / 168

Results

21 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Berkeley Function Calling Leaderboard v4Agentic66.6%11 / 26
American Invitational Mathematics Examination 2025Math86.5%25 / 53
Instruction-Following EvalInstruction86.7%35 / 70
MATH-500 Problem SetMath94.6%17 / 27
Critical Physics TasksReasoning0.3%137 / 214
Artificial Analysis Long Context ReasoningReasoning59.0%143 / 217
Artificial Analysis GPQA DiamondKnowledge70.2%132 / 199
Artificial Analysis Humanity's Last ExamKnowledge8.9%151 / 222
GDPval-AA normalizedAgentic10.4%92 / 133
LiveCodeBench v6Coding69.1%42 / 59
Massive Multitask Language Understanding ProfessionalKnowledge70.8%55 / 75
Instruction Following BenchmarkInstruction66.3%42 / 57
AIME 2026Math86.5%28 / 34
GPQA Diamond · BenchLMKnowledge70.2%72 / 84
MMLU-ReduxKnowledge84.7%17 / 19
Software Engineering Benchmark VerifiedCoding46.4%104 / 114
Harvard-MIT Mathematics Tournament February 2026Math63.8%25 / 27
Artificial Analysis SciCodeCoding26.3%106 / 114
SuperGPQA: Scaling LLM Evaluation Across 285 Graduate DisciplinesKnowledge40.8%21 / 22
LongBench v2Reasoning43.7%24 / 25
Humanity's Last ExamKnowledge8.9%74 / 77
Scientific Code BenchmarkCoding26.3%34 / 35
SWE-bench ProCoding14.4%83 / 83
Terminal-Bench 2.1 · BenchLMAgentic8.6%86 / 86

Sources

BenchLMCC BY-NC 4.0

History

5 releases
Index37.6 → 37.1
23 Oct 202612 Nov 2026

Questions

How good is MiniCPM5-2B?
Its index is 37.1 out of 100. That is place 226 of 283 ranked models.
What is MiniCPM5-2B best at?
Instruction: 44.4 out of 100, place 39 of 63.
What is the context window of MiniCPM5-2B?
131K tokens.
Is MiniCPM5-2B open source?
Yes. Its weights are published.
When was MiniCPM5-2B released?
September 2026.

Badge

MiniCPM5-2B on Inference360: place 226 of 283, index 37.1
HTML
<a href="https://inference360.ai/models/minicpm5-2b"><img src="https://inference360.ai/badges/minicpm5-2b.svg" alt="MiniCPM5-2B on Inference360: place 226 of 283, index 37.1" height="20"></a>
Markdown
[![MiniCPM5-2B on Inference360: place 226 of 283, index 37.1](https://inference360.ai/badges/minicpm5-2b.svg)](https://inference360.ai/models/minicpm5-2b)

Compare with

DeepSeek V3 032437.1Nemotron 3.5 Lightning 30B A3B NVFP437.0Ling 2.6 Flash37.0Granite 4.2 8B37.6MiniCPM5-1B15.5

More from OpenBMB

All models
MiniCPM5-1B15.5MiniCPM-o 2.6Unranked