OpenBMB Releases MiniCPM5-2B-GGUF for On-Device AI
At a Glance
| Item | Value |
|---|---|
| Repository | openbmb/MiniCPM5-2B-GGUF |
| Published | 2026-09-05 |
| License | apache-2.0 |
| Formats | GGUF |
| Paper | arXiv:2506.07900, arXiv:2602.09003 |
| Source type | Primary source (the publisher itself) |
Values determined by this site’s code at collection time. Dates are JST.
Overview
OpenBMB has released “MiniCPM5-2B-GGUF", a 2B-class open-weight model designed for on-device and resource-constrained environments. This model is the second in the series following MiniCPM5-1B and adopts the standard LlamaForCausalLM architecture. It is designed targeting local assistants, coding agents, tool-use workflows, and reasoning scenarios.
Specifications
- Parameters: 2,516,756,480
- Non-embedding parameters: 1,981,982,720
- Layers: 42
- Attention heads (GQA): 16 for Q, 2 for KV
- Context length: 131,072
- License: apache-2.0
Performance
In the evaluation results of the model card, MiniCPM5-2B is compared with open-source models of similar scale such as LFM2.5-2.6B, Qwen3.5-2B, Gemma-4-E4B-it, as well as larger models including Qwen3.5-4B, granite-4.2-3B, Nemotron-3-Nano-4B, Gemma-4-E4B-it, and LFM2.5-8B-A1B. Here, the comparison targets are narrowed down to five: MiniCPM5-2B, LFM2.5-2.6B, Qwen3.5-2B, Qwen3.5-4B, and granite-4.2-3B.
→ Scroll horizontally to see all columns
| MiniCPM5-2B | 2B-class Models / LFM2.5-2.6B | 2B-class Models / Qwen3.5-2B | 4B-class Models / Qwen3.5-4B | 4B-class Models / granite-4.2-3B | |
|---|---|---|---|---|---|
| Average | 53.9 | 33.2 | 28.0 | 51.1 | 42.7 |
| Code Reasoning | |||||
| LiveCodeBench v6 | 69.1 | 42.1 | 20.2 | 56.4 | 58.9 |
| LCB-Pro 25Q2 (Easy) | 68.0 | 30.9 | 10.3 | 58.3 | 54.6 |
| LCB-Pro 25Q2 (Medium) | 17.5 | 0.0 | 0.0 | 7.0 | 5.3 |
| OJBench | 32.5 | 11.2 | 2.6 | 24.8 | 21.8 |
| SciCode (wbg) | 26.3 † | 14.2 † | 2.8 † | 16.1 † | 24.9 † |
| Math Reasoning | |||||
| AIME 2025 | 86.5 | 41.9 | 29.6 | 78.8 | 79.4 |
| AIME 2026 | 86.5 | 45.2 | 29.0 | 82.7 | 83.5 |
| HMMT Feb 2026 | 63.8 | 33.7 | 20.5 | 64.0 | 60.8 |
| MATH-500 | 94.6 | 89.6 | 85.8 | 99.0 | 97.0 |
| Instruction Following | |||||
| IFBench | 66.3 | 59.0 | 46.0 | 59.0 | 73.0 |
| IFEval | 86.7 | 93.4 | 77.5 | 90.2 | 93.7 |
| Multi-IF | 71.8 | 76.8 | 57.1 | 73.6 | 75.9 |
| General Knowledge | |||||
| MMLU-Pro | 70.8 | 65.2 | 64.3 | 78.0 | 65.8 |
| MMLU-Redux | 84.7 | 80.0 | 80.0 | 88.7 | 78.9 |
| HLE | 8.9 † | 6.2 † | 2.6 † | 9.9 † | 6.6 † |
| GPQA-Diamond | 70.2 † | 55.8 † | 45.6 † | 77.1 † | 55.9 † |
| SuperGPQA | 40.8 | 26.2 | 38.6 | 52.8 | 39.9 |
| Long Context | |||||
| AA-LCR | 59.0 † | 5.3 † | 28.7 † | 61.0 † | 24.3 † |
| NoLiMa | 68.1 | 0.7 | 17.1 | 43.5 | 5.1 |
| LongBenchPro | 44.8 | 23.7 | 8.2 | 58.4 | 34.8 |
| LongBench v2 | 43.7 | 30.3 | 24.9 | 47.3 | 36.0 |
| Tool Use | |||||
| τ³-Bench Banking | 20.8 † | 7.2 † | 2.1 | 6.8 † | 5.6 † |
| τ²-Bench Telecom | 97.1 | 90.4 | 69.0 † | 92.1 † | 40.9 |
| BFCL v4 | 66.6 | 61.1 | 43.6 | 56.8 | 52.2 |
| Coding Agent | |||||
| SWE-bench Verified | 46.4 | 6.0 | 5.0 | 33.6 | 36.8 |
| SWE-bench Pro | 14.4 | 0.6 | 0.8 | 28.2 | 12.3 |
| Terminal-Bench v2.1 | 8.6 † | 4.5 † | 3.0 † | 25.8 † | 13.9 † |
| Search Agent | |||||
| BrowseComp-ZH | 43.5 | 9.8 | 18.2 | 39.6 | 21.1 |
| BrowseComp Top100 | 39.7 | 13.7 | 19.3 | 33.3 | 19.0 |
| GAIA Text-103 | 88.7 | 49.5 | 47.9 | 78.6 | 57.3 |
| General Agent | |||||
| GDPval-AA v2 | 19.6 † | 4.5 | 0.0 | 11.7 | 0.0 † |
| Claw-Gym | 59.2 | 19.3 | 25.5 | 51.6 | 60.0 |
| WildClaw | 23.9 | 10.2 | 9.2 | 17.0 | 20.0 |
| QwenClaw | 42.9 | 19.3 | 18.2 | 37.1 | 36.4 |
According to measurements by the publishers, MiniCPM5-2B recorded an average score of 53.9, demonstrating superiority compared to 2B-class models of similar scale. It achieved high scores in categories such as code reasoning (e.g., LiveCodeBench v6), competitive mathematics (e.g., AIME 2025/2026), tool use (e.g., τ²-Bench Telecom), and research-oriented agents (e.g., GAIA Text-103). On the other hand, it scores lower than some models in certain instruction-following metrics (e.g., IFEval).
Strengths and Use Cases
The model card lists local assistants, coding agents, tool-use workflows, and reasoning scenarios as its primary use cases. Supporting a context length of 131,072, it is designed with strengths in long-context understanding, code and math reasoning, and agent tasks.
Hardware Requirements
Estimated requirements (calculated by Local Model Watch)
| Your VRAM | Quantization | File size | Est. memory needed |
|---|---|---|---|
| 8GB (RTX 4060 / 3060 Ti, etc.) | F16 | 4.7GB | 5.6GB |
Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference. What the quantization names mean: glossary.
How to Get It
Distributed in GGUF format for llama.cpp, Ollama, and LM Studio.