Aiqre

Model library / openbmb

MiniCPM5-2B-Base

MiniCPM5-2B is a 2.5 billion-parameter dense Llama-style causal model from openbmb for on-device, resource-constrained deployment.

2.5B

Parameters in total

524,288

Context length in tokens, as published

5 GiB

Weights at the released precision

apache-2.0

Licence declared on the repository

Overview

What this model is.

MiniCPM5-2B is a dense 2.5 billion-parameter Llama-style causal language model released by openbmb. It is intended for on-device, local deployment and resource-constrained scenarios, extending the MiniCPM5 series. The model is trained with a high-quality web dataset and specialised code, agent and reinforcement-learning data to improve coding, mathematics and long-context tasks.

The model occupies roughly 5 GB in bfloat16 format, reflecting its 2.5 billion parameters. It supports a context window of 524 288 tokens, enabling very long inputs. Distribution is under the Apache-2.0 licence, allowing commercial use. It uses the LlamaForCausalLM architecture without expert layers, and runs in bfloat16 precision.

Specification

The published shape.

Repositoryopenbmb/MiniCPM5-2B-Base
Publisheropenbmb
Published27 August 2026
ArchitectureLlamaForCausalLM
Model typellama
Parameters2,516,756,480
Layers42
Hidden size2,048
Attention heads16, 2 key/value heads
Context length524,288 tokens
Vocabulary130,560 tokens
Weight formatbfloat16
Licenceapache-2.0

Memory

How much VRAM the weights need.

Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.

PrecisionWeightsNotes
bf164.7 GiBFull precision as released
fp82.3 GiB8-bit, near-lossless on most models
int41.2 GiB4-bit, smallest footprint

The cache costs 42 KiB per token at 16-bit, so the full 524,288-token context of one request needs about 21.0 GiB. Concurrency multiplies that number, not the weights.

Hardware

Which card runs it.

At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.

GPUCards neededFree for cacheContext it holds
RTX 4090 24GB 1 14.4 GiB about 359,815 tokens
RTX 5090 32GB 1 21.6 GiB the full 524,288 tokens
L40S 48GB 1 36.0 GiB the full 524,288 tokens
A100 80GB 1 64.8 GiB the full 524,288 tokens
H100 80GB 1 64.8 GiB the full 524,288 tokens
RTX PRO 6000 96GB 1 79.2 GiB the full 524,288 tokens
H200 141GB 1 119.7 GiB the full 524,288 tokens

Serving

Running it yourself.

A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.

vllm serve openbmb/MiniCPM5-2B-Base \
  --max-model-len 356352 \
  --gpu-memory-utilization 0.90

Questions

The things people ask about this model.

How much GPU memory does MiniCPM5-2B-Base need?

The weights take about 5 GiB at the released precision. That fits on one RTX 4090 24GB, which leaves roughly 14 GiB for the key-value cache.

What licence does MiniCPM5-2B-Base use?

The repository declares apache-2.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.

How long a context does it support?

The configuration allows 524,288 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.

Can I use it through an API instead of hosting it?

This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.

Also in the library

Related models.

openbmb/MiniCPM5-2B

MiniCPM5-2B is a 2-billion-parameter dense transformer aimed at on-device and resource-constrained deployments.

openbmb/MiniCPM5-2B-Midtrain

MiniCPM5-2B is a 2.5-billion-parameter dense Llama-style causal language model released by openbmb.

openbmb/MiniCPM5-2B-SFT

MiniCPM5-2B-SFT is a 2.5 billion-parameter dense Llama-style transformer released by OpenBMB for on-device, resource-constrained …

TokenRhythm/NeoHorse-1-4B

NeoHorse-1-4B is a 4.2B open-weight model published by TokenRhythm on Hugging Face. It uses the Qwen3_5ForCausalLM architecture w…

Hosting

Want this model on a dedicated EU GPU?

Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.