Leaderboard / OpenBMB (MiniCPM)

MiniCPM-V 4.7 35B-A3B

Name: MiniCPM-V 4.7 35B-A3B
Author: OpenBMB (MiniCPM)

OpenBMB (MiniCPM) · MiniCPM-V · released 2026-10-07 · data checked 2026-11-02

MiniCPM-V 4.7 35B-A3B is a 35.2B-parameter Mixture-of-Experts model (3B active per token) from OpenBMB (MiniCPM) (China), released 2026-10-07 under the Apache-2.0 license. It scores 54.5 on the Small LM Index, #24 of 34 medium models. It needs about 22.7 GB of memory at 4-bit with an 8K-token context.

Vision-language MoE. Text scores are from the card in instruct (non-thinking) mode; its many vision results (MMBench 91.7, MMMU 80.2) use benchmarks this site does not track.

Small LM Index

54.5

#24 of 34 scored medium models

Likely 50.0–58.9 · from 3 of 6 core benchmarks

Parameters

35.2B

3B active · MoE

Price per 1M tokens

—

No hosted API on OpenRouter; self-host

Context window

262K

tokens

Benchmarks

Each bar shows this model's score. The tick marks the best medium model on that benchmark. Sources: 6 reported by OpenBMB (MiniCPM). How we label sources.

MMLU-Pro Lab-reported Knowledge · Instruct mode (all text scores)83.5

Class best: Qwen3.6 27B 86.2

AIME Lab-reported Math · AIME 2026, avg@1676.5

Class best: Ternary Bonsai 2 27B 95.8

LiveCodeBench Lab-reported Coding · LiveCodeBench v656.6

Class best: Qwen3.8 27B 90.3

τ²-Bench Lab-reported Agentic84.7

IFBench Lab-reported Instruction following47.7

Class best: Sage-A1.5 81.6

BFCL Lab-reported Agentic · BFCL v460.1

Class best: Ternary Bonsai 2 27B 74.9

Not reported: estimated

What the index model expects from this model's other scores. Not a measurement.

GPQA Diamond≈ 68

Humanity's Last Exam≈ 8

SWE-bench Verified≈ 37

Specifications

Lab: OpenBMB (MiniCPM) (CN · China)
Based on: Qwen3.5 35B-A3B (fine-tune or distillation)
Architecture: Mixture-of-Experts
Total parameters: 35.2B
Active parameters: 3B
Size class: Medium
Context window: 262K tokens
Input modalities: text, image, video
Reasoning: Hybrid (thinking on/off)
License: Apache-2.0
Openness: Open weights
Memory to run: ~22.7 GB at 4-bit, 8K context (more sizes)
Hugging Face: 490.1K downloads (30 days) · 1.1K likes

Weights on Hugging Face ↗

Compare with

MiniCPM-V 4.7 35B-A3B vs Qwen3.6 35B-A3B (35B, index 72.2)
MiniCPM-V 4.7 35B-A3B vs Qwen3.5 35B-A3B (35B, index 69.9)
MiniCPM-V 4.7 35B-A3B vs Edge0 35B-A3B (preview) (34.7B, index 62.8)
MiniCPM-V 4.7 35B-A3B vs Seed-OSS 36B Instruct (36.2B, index 60.3)
MiniCPM-V 4.7 35B-A3B vs Hermes 4.3 36B (36.2B, index 51.4)

Sources

https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B

Something wrong here? Report it.

Run it locally

No GGUF build from a well-known publisher yet.

vLLM (GPU server)

vllm serve openbmb/MiniCPM-V-4.7-35B-A3B

Full-precision weights, for a server GPU.

Memory needed: weights, context cache and runtime buffers.

Quantization	4K context	32K context	256K context
16-bit F16	72 GB	72 GB	77 GB
8-bit Q8_0	39 GB	39 GB	44 GB
4-bit Q4_K_M	23 GB	23 GB	28 GB

Check it against your GPU, Mac or phone →