OpenBMB

2025•OpenBMB

VoxCPM v0.5B

VoxCPM is an open-source, tokenizer-free text-to-speech (TTS) engine designed for context-aware speech generation and true-to-life voice cloning. It utilizes hierarchical language modeling to synthesize high-quality continuous speech representations directly from input text, supporting real-time streaming output of generated audio segments.

Tokenizer-free end-to-end architecture

Zero-shot voice cloning with 3-10 seconds of reference audio

High efficiency with real-time inference on consumer GPUs

PricingFree

Version0.5B

2026•OpenBMB

MiniCPM-SALA v1.0

Open Source

LLMs

MiniCPM-SALA is a large-scale hybrid language model developed by OpenBMB, integrating sparse and linear attention mechanisms to efficiently handle long-context modeling. Released on February 11, 2026, it is the first model to effectively combine sparse and linear attention for million-token context modeling. ([huggingface.co](https://huggingface.co/openbmb/MiniCPM-SALA?utm_source=openai))

Hybrid architecture combining 25% sparse attention and 75% linear attention

Achieves 3.5× inference speedup over dense baselines

Supports context lengths up to 1 million tokens

PricingOpen-source and freely available

Version1.0

2026•openbmb

UltraData-Math v1.0

Open Source

UltraData-Math is a large-scale, high-quality mathematical pre-training dataset designed to enhance mathematical reasoning in large language models (LLMs). It comprises over 290 billion tokens across three progressive tiers: L1 (170.5B tokens from web corpus), L2 (33.7B tokens from quality-selected data), and L3 (88B tokens from multi-format refined data). This dataset has been utilized in the mathematical pre-training of the MiniCPM Series models.

Comprehensive dataset with over 290 billion tokens

Three progressive tiers (L1, L2, L3) for systematic enhancement

Applied in pre-training MiniCPM Series models

PricingFree

Version1.0