OpenBMB

MiniCPM-V 4.7 35B-A3B

OpenBMB

MiniCPM-V 4.7 35B-A3B is a public OpenBMB vision-language checkpoint uploaded on October 6. Its official repository contains 16 downloadable BF16 weight shards totaling approximately 35.21B parameters, a hybrid linear/full-attention mixture-of-experts language configuration, and image/video processing metadata. The configuration permits 262,144 positions, but no model card or validated context specification was published in the inspected repository, so no supported context limit is asserted here. Text and visual inputs produce language output through the conditional-generation architecture. The license, training ancestry, evaluated capabilities, and benchmark results remain undisclosed at the audit time; a public weight upload does not supply missing usage rights.

MathForm-8B

OpenBMB

MathForm-8B is OpenBMB’s language-to-code model for translating natural-language mathematical statements into Lean 4. It fine-tunes Qwen3-8B on FormalVerse with supervised learning followed by reinforcement learning using compilation and semantic-consistency feedback. The research framework retrieves Mathlib definitions and revises generated formalizations using diagnostics, supporting more faithful mathematical translation. Apache-licensed weights, approximately 367K verified examples, and evaluation code accompany the August 14 paper. This is a named mathematical language specialization; the model is distinct from the surrounding retrieval and verification pipeline. Successful compilation alone does not establish that a formal statement preserves the intended proposition.

UltraX

OpenBMB

UltraX is OpenBMB’s specialized 0.6B language-model release for refining pre-training text through structured editing function calls. It generates insertions, deletions, and replacements rather than rewriting an entire document, and a deterministic executor applies validated edits. Full-parameter supervised training uses program supervision built from language-model refinements and alignment procedures. The July 9 paper precedes the July 13 public code-and-model announcement. Apache-licensed refinement checkpoints and data are available, with no-instruction and preservation-weighted variants grouped here. Its named language-editing specialization qualifies separately from generic data tools or probability-only classifiers.

MiniCPM5

OpenBMB

MiniCPM5 is OpenBMB’s fifth-generation dense language-model family for local and resource-constrained use. The 1.08B model was announced May 19, followed by a 2.52B model on September 7. Both support a documented 128K context and target reasoning, coding, tool use, writing, and instruction following. The training recipe combines high-quality data, instruction tuning, and complementary reinforcement-learning signals. Apache-licensed base, intermediate, supervised, and final checkpoints are grouped as variants, alongside routine deployment formats. Small parameter counts improve local practicality but do not eliminate errors or the need for application-specific evaluation.

MiniCPM-V 4.6

OpenBMB

MiniCPM-V 4.6 is OpenBMB’s compact 1.3B vision-language model for efficient device deployment. It combines Qwen3.5-0.8B with SigLIP2-400M and mixed 4x/16x visual-token compression, accepting text, images, and video and producing language answers. The release focuses on reducing visual encoding work while preserving multi-image and video understanding. Apache-licensed weights, ordinary and Thinking checkpoints, a demo, and a later May 17 API are available. This page does not equate the backbone configuration maximum with validated long-context visual use. Throughput comparisons are provider measurements tied to particular input and serving conditions.

MiniCPM-SALA

OpenBMB

MiniCPM-SALA is OpenBMB’s 9.48B language model combining sparse and linear attention for million-token contexts. One quarter of layers use InfLLM-v2 sparse attention and three quarters use Lightning Attention, with Hybrid Positional Embedding supporting length generalization beyond short training sequences. Apache-licensed weights and inference examples target long-context local language tasks with reduced cache overhead. The official model card documents deployment at up to 1M tokens, extending beyond its static 512K configuration value. Speedup claims depend on hardware, sequence length, and implementation rather than applying uniformly to every prompt.

MiniCPM-o 4.5

OpenBMB

MiniCPM-o 4.5 is OpenBMB’s end-to-end 9B multimodal conversation model, combining Qwen3-8B, SigLIP2, Whisper-medium, and CosyVoice2 components. It processes text, images, continuous video, and audio while generating concurrent text and speech. Full-duplex streaming supports speaking while listening and proactive responses to a live scene. Thinking and instruction modes balance reasoning with responsiveness. Apache-licensed weights and local demo code were released February 3, before the later technical report and hosted API. It qualifies as a language-capable multimodal model, including substantial visual and spoken question answering rather than transcription or synthesis alone.

MiniCPM4.1

OpenBMB

MiniCPM4.1 is OpenBMB’s 8B language model with trainable sparse attention and hybrid reasoning. InfLLM-v2 enables efficient attention to long text, while fusion-thinking post-training supports both deliberative and direct answers. The native context is 64K; the model card separately validates 128K with LongRoPE scaling, which is recorded as an extension rather than the default. Apache-licensed weights and local deployment code are available. Its efficiency focus suits private assistants and coding applications on devices or modest servers. Sparse attention settings, reasoning mode, and context scaling affect both quality and runtime behavior.

MiniCPM-V 4.5

OpenBMB

MiniCPM-V 4.5 is OpenBMB’s open-weight vision-language model based on Qwen3-8B and SigLIP2-400M. It supports text, single or multiple images, and video understanding with language output. High-resolution image processing and compressed video tokens target document questions, visual reasoning, and longer visual sequences on local devices. It offers thinking and instruction behavior alongside multilingual capabilities spanning more than 30 languages. Apache-licensed weights, inference examples, and a demo are available. Reported OCR and document benchmarks reflect particular test conditions and do not make extracted information automatically reliable.

MiniCPM4

OpenBMB

MiniCPM4 is OpenBMB’s efficient local language-model generation, with 8B and 0.5B checkpoints. It combines training-data improvements with sparse-attention and deployment optimizations intended for devices and constrained servers. Apache-licensed weights, code, and a technical report support local research and customization. This family page also groups BitCPM4 and the later BitCPM-CANN low-bit training variants derived from MiniCPM backbones. CANN checkpoints demonstrate ternary quantization-aware training on Huawei Ascend; their released pseudo-quantized floating-point storage does not itself provide a packed low-memory inference representation. The separately named MiniCPM4.1 update has its own page.