// Open-model VRAM table · family
MiniCPM / VoxCPM: local deployment & VRAM
23 open-weight models from OpenBMB (75 official repos incl. quantised builds). At 4-bit they range from 0.9 GB (MiniCPM4-0.5B) to 24.2 GB (MiniCPM-V-4.7-35B-A3B).
official orgs: openbmblatest 2026-10-07memory = estimate
01Text LLM (8)
| Model | Params | Context | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|---|
| MiniCPM5-2B | 2.52B | 128K | 6.4 GB | 2.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-09-06 |
| MiniCPM5-1B | 1.08B | 128K | 3.1 GB | 1.4 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-05-21 |
| MiniCPM-SALA | 9.48B | 512K | 21.6 GB | 7.1 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-02-11 |
| MiniCPM4.1-8B | 8.19B | 64K | 18.8 GB | 6.2 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-09-02 |
| MiniCPM4-0.5B | 434M | 32K | 1.6 GB | 0.9 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-06-05 |
| MiniCPM4-8B | 8.19B | 32K | 18.8 GB | 6.2 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-06-05 |
| MiniCPM4-MCP | 8.19B | 32K | 18.8 GB | 6.2 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-06-05 |
| MiniCPM3-4B | 4B | 32K | 9.6 GB | 3.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2024-09-03 |
02Vision-language (8)
| Model | Params | Context | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|---|
| MiniCPM-V-4.7-35B-A3B | 35.2B / 3B | 256K | 78.1 GB | 24.2 GB | RTX 5090 32GB | Mac 64GB | Commercial OK | 2026-10-07 |
| MiniCPM-V-4.7-1B | 1.55B | 256K | 4.0 GB | 1.6 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-10-06 |
| MiniCPM-V-4.6-Thinking | 1.3B | 256K | 3.5 GB | 1.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-05-08 |
| MiniCPM-V-4.6 | 1.3B | 256K | 3.5 GB | 1.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-04-13 |
| MiniCPM-V-4_5 | 8.7B | 40K | 20.8 GB | 7.5 GB | 12GB GPU | Mac 16GB | Commercial OK | 2025-08-24 |
| MiniCPM-V-4 | 4.06B | 32K | 9.7 GB | 3.5 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-07-12 |
| MiniCPM-V-2_6 | 8.1B | — | 18.3 GB | 5.9 GB | 8GB GPU | Mac 16GB | Licence not stated | 2024-08-04 |
| MiniCPM-Llama3-V-2_5 | 8.54B | 8K | 20.4 GB | 7.3 GB | 12GB GPU | Mac 16GB | Licence not stated | 2024-05-19 |
03Omni / audio LLM (2)
| Model | Params | Context | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|---|
| MiniCPM-o-4_5 | 9.37B | 40K | 22.3 GB | 8.0 GB | 12GB GPU | Mac 16GB | Commercial OK | 2026-02-03 |
| MiniCPM-o-2_6 | 8.68B | 32K | 20.1 GB | 6.8 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-01-12 |
04Embedding (2)
| Model | Params | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|
| MiniCPM-Embedding-Light | 434M | 1.5 GB | 0.8 GB | 8GB GPU | Mac 16GB | Licence not stated | 2025-01-17 |
| MiniCPM-Embedding | 2.73B | 6.5 GB | 2.3 GB | 8GB GPU | Mac 16GB | Licence not stated | 2024-09-04 |
05Text-to-speech / audio (3)
| Model | Params | BF16 | 4-bit | Smallest GPU (4-bit) | Mac (4-bit) | Licence | Released |
|---|---|---|---|---|---|---|---|
| VoxCPM2 | 2.29B | 5.5 GB | 2.0 GB | 8GB GPU | Mac 16GB | Commercial OK | 2026-04-03 |
| VoxCPM1.5 | 802M | 2.3 GB | 1.0 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-12-05 |
| VoxCPM-0.5B | 500M | 1.6 GB | 0.8 GB | 8GB GPU | Mac 16GB | Commercial OK | 2025-09-16 |
06Other families
Qwen · IBM Granite · NVIDIA Nemotron · Gemma · Mistral / Magistral / Devstral · GLM / CogVideoX (Z.ai) · DeepSeek · Ai2 OLMo / Molmo · Ant Ling / Ring / Ming · Liquid LFM · Tencent Hunyuan / Hy · Phi · Llama · Meituan LongCat · Jina embeddings & rerankers · Wan · Kimi · Cohere Command / Aya · Xiaomi MiMo · FLUX · Microsoft VibeVoice / Fara / Harrier · InternLM / Intern-S · StepFun Step · MiniMax · Baidu ERNIE · NVIDIA Parakeet / Canary speech · Stable Diffusion / Stable Audio · Yi (01.AI) · ByteDance Seed · SmolLM / SmolVLM · BGE embeddings & rerankers · LTX Video · FunAudioLLM speech · gpt-oss · poolside Laguna · Whisper · Baichuan · Thinking Machines Inkling · xAI Grok · rednote dots
Totals include 10% + 0.5GB overhead and one 8K-token KV cache for text models. Estimates only. Method →