Model
Model explorer

MiniCPM-Llama3-V 2.5

OPEN
OpenBMB · MiniCPM-V family · released May 20, 2024

First end-side 8B MLLM to reach GPT-4V level, surpassing Gemini Pro and Claude 3 on OpenCompass (65.1 avg).

ReasoningCodingVisionFunction callingTool useAgentic
634.2
Elo · unrated
Parameters
8B
Active params
8B (dense)
Context
8K tokens
Architecture
SigLip-400M vision encoder + Llama3-8B-Instruct LLM
License
MiniCPM Model License (custom; free for research and, after registration, commercial use) + Apache-2.0 code
Languages
30+
API price (in/out)
No hosted API
Modalities
text · vision
Benchmark results
Bar shows position within the tracked field; marker = field best
DocVQAVision84.8%#40
best: Qwen2-VL-72B · 96.5%
MathVistaVision54.3%#49
best: Seed 2.1 Pro · 90.7%
MMBench (Chinese)Vision74.2%#8
best: ERNIE 4.5 VL 424B-A47B · 90.9%
MMBench (English)Vision77.2%#18
best: Qwen3.5-397B-A17B · 93.7%
MMMUVision45.8%#101
best: Claude Fable 5 · 89.3%
OCRBenchVision725#12
best: InternVL3-78B · 906
TextVQAVision76.6%#16
best: Molmo 2 8B · 85.7%
Run it locally
VRAM @ Q4
6 GB
VRAM @ FP16
16 GB
Fits on (Q4)
RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Throughput data unavailable.
Quantizations
GGUF Q4_K_M · GGUF Q5_K_S · GGUF Q2_K
MiniCPM-V family
Elo progression across releases
MiniCPM-Llama3-V 2.5May 2024634.2MiniCPM-V 2.6Aug 2024771.6
Resources
HFopenbmb/MiniCPM-Llama3-V-2_5GHOpenBMB/MiniCPM-V
API price weights · each benchmark row carries its own source badge (see methodology)