NVIDIA Nemotron 3.5 Super VL 对比 MiniCPM-V 4.7 35B-A3B：口袋中的 3 GB 对阵机架中的 4× H100

One of these runs on an iPhone. The other runs on four H100s. That is the whole comparison, and everything else is detail. MiniCPM-V 4.7 35B-A3B is OpenBMB's first Mixture-of-Experts entry in the MiniCPM-V line, and the model card's claim is specific enough to check: more than 13 tokens per second on an iPhone 17 Pro Max using roughly 3 GB of memory, with the run described as smooth rather than as a demo. N​VIDIA N​emotron 3.5 Super VL is a 120B-total, 12B-active vision-language model whose card puts the deployment envelope at 4× H100 or 2× GB200. Both are open weights. Both are vision-language models. They are not options for the same decision.

无论如何，它们值得比较的原因是：它们在机制上一致，而在规模上不同。两者都是 Mixture-of-Experts 模型，都是为了用总规模换取每 token 成本而构建的。两者都在语言模型必须思考它们之前压缩视觉 token。两者都押注：视觉-语言工作的未来是许多模型分布在许多地方，而不是一个大模型藏在一个 API 后面。只不过，它们在硬件上的下注相差三个数量级。

活动参数巧合

N​VIDIA N​emotron 3.5 Super VL is 120B total with 12B active per token — a Mamba-2–Transformer hybrid MoE with a C-RADIO vision encoder and 512 routed experts per MoE layer, holding up to 1M tokens of context. MiniCPM-V 4.7 35B-A3B is 35B total with 3B active. Different by a factor of four in total size and four in active compute, but the same architectural instinct: keep the parameter count high for knowledge and the per-token cost low for speed.

OpenBMB 对该问题的贡献是 Lookahead Expert Scheduling，它让一个 3B 激活参数的模型能够将其专家保持常驻，而不会反复冲击内存总线——这正是它能在 3 GB 中运行、而不需要一台工作站的原因。其实现位于一个分支 llama.cpp-omni 中，随附的说法也很谨慎：首个能在设备端流畅运行且不牺牲性能的开源 MoE 模型。至于“不牺牲性能”在面对 120B 模型时是否成立，并不是这张模型卡要说明的事情，而且也不应该由它说明——一个能在手机上运行的模型，在准确率上并不与一个需要四块 H100 的模型竞争。它竞争的是能否存在本身。

• Total / active parameters — N​emotron 3.5 Super VL: 120B / 12B. MiniCPM-V 4.7 35B-A3B: 35B / 3B.

• Deployment target — N​emotron 3.5 Super VL: 4× H100 or 2× GB200, served via vLLM or SGLang. MiniCPM-V 4.7 35B-A3B: phone, laptop, edge box — iOS, Android and HarmonyOS are all named.

• Memory at the smallest end — N​emotron 3.5 Super VL: not applicable. MiniCPM-V 4.7 35B-A3B: about 3 GB.

• Context — N​emotron 3.5 Super VL: up to 1M tokens, video up to two minutes. MiniCPM-V 4.7 35B-A3B: not the model's selling point; the card leads on footprint and speed.

• Licence — N​emotron 3.5 Super VL: OpenMDW-1.1. MiniCPM-V 4.7 35B-A3B: Apache-2.0.

对“token 很贵”的两种不同回答

两支团队都着手解决同一个瓶颈——一张图片所耗费的视觉 token 数量——却抵达了不同的境地。

MiniCPM-V 4.7 35B-A3B 在视觉编码器内部完成这一过程：先采用源自 LLaVA-UHD v4 的早期压缩，然后根据图像的实际需要，混合使用 4× 和 16× 的视觉 token 压缩。精细细节走成本更高的路径，而一张收据则走低成本路径。压缩发生在语言模型介入之前，这正是端侧指标得以实现的原因。

N​emotron 3.5 Super VL does not lead with token compression at all. Its answer is context: 1M tokens, so you can afford to keep the detail rather than throw it away, and a hybrid Mamba-2 layer to keep the cost of a very long sequence from exploding. The model's own video path is configured generously by default — 2 FPS and 256 frames, dropping to 1 FPS and 128 frames at 1080p — which is a sampling answer, not a compression answer.

这是两种互不相容的理念，但对各自的部署场景而言都是正确的。如果你只有 3 GB，那就压缩。如果你有 4× H100，那就保留细节，为上下文买单。

数字能告诉你什么，不能告诉你什么

以下是双方诚实的立场。

N​emotron 3.5 Super VL's benchmark table is broad and vendor-collected through NeMo Gym: GPQA-Diamond 89.52, SWE-bench Verified 71.10, MMLongBench-Doc 62.82, OCRBench v2 at 68.12 English and 59.15 Chinese, CharXiv reasoning 80.82, V* 85.60. N​VIDIA chose the baselines. None of it has been independently reproduced.

MiniCPM-V 4.7 35B-A3B 的模型卡以厂商图表而非数值表格展示其基准测试结果，因此没有可引用的数字，也没有可供查看的基线列表。模型卡中最强的具体说法是比较性且方向性的——在 Instruct 模式下，相比 Qwen3.6 35B-A3B，以更少推理时间获得更高总体得分——这说的是其自身参数量级内的效率，而不是相对于一个参数量为其十二倍的模型的能力。

所以，这两者在基准测试数据上的对比表并不存在，也无法被构建出来。任何拿出这样一张表的人，不过是在把厂商的图表和厂商的表格拼凑在一起，然后把结果称作一次测量。

比准确性决定更多项目成败的部署细节

MiniCPM-V 4.7 35B-A3B 在发布首日即支持 Rockchip、Houmo、Intel、AMD 和 Arm 芯片，同时兼容 SGLang、vLLM、llama.cpp 和 Ollama，并提供 GGUF、BNB、AWQ、GPTQ 和 FP8 量化格式。它还与 SWIFT 和 LLaMA-Factory 集成，用于微调。这份清单才是真正的产品。一个能运行在你已经购买的加速器上、采用你的运行时已支持的量化格式的模型，才是你本季度就能交付的模型。

N​emotron 3.5 Super VL's runtime story is narrower and deeper: vLLM at vllm/vllm-openai:v0.32.0 and SGLang at lmsysorg/sglang:v0.5.20, with EAGLE speculative decoding at topk 1 and Multi-Token Prediction built into the architecture. Two runtimes, both first-class, both specified to the exact version. That is what a datacenter deployment wants — a validated path, not a menu.

If you are choosing hardware to fit the model, take MiniCPM-V 4.7 35B-A3B. If you are choosing the model to fit hardware you already have, take N​emotron 3.5 Super VL and follow its card's versions exactly.

同时运行两者的理由

The pattern that shows up in real products is not a choice between these two. It is a small VLM on the device handling the work that has to be instant and local — reading a form, describing a scene, filtering out the frames nobody cares about — and a large VLM in the datacenter handling the work that has to be right. MiniCPM-V 4.7 35B-A3B is explicitly designed for the first role. N​emotron 3.5 Super VL is designed for the second.

That architecture has a cost, and the cost is that you are now running two model stacks and a text stack, with three sets of credentials and three sets of failure modes. OrcaRouter is the part that removes the second problem: 207 models behind one key at provider list price with zero markup, so a vendor price change lands on your bill the same day, and automatic failover when an upstream degrades. Neither N​VIDIA N​emotron 3.5 Super VL nor MiniCPM-V 4.7 35B-A3B is in that catalogue — both are open weights you host yourself, which is the point of choosing either. What one key gets you is the surrounding text and code models, so the pipeline does not need a new contract every time you add a stage.

哪一个，由像素所在的位置决定

如果推理必须发生在数据所在之处——在手机上、在车辆中、在工厂车间里、在图像不能离开大楼的诊所内——或者如果你的用户处于离线状态，又或者托管式 VLM 的每次调用成本正是压垮单位经济效益的元凶，那就选 MiniCPM-V 4.7 35B-A3B。它的许可证是 Apache-2.0，量化选项丰富，芯片支持列表也刻意做得很长。

Take N​VIDIA N​emotron 3.5 Super VL if the work is long-context and reasoning-heavy: multi-page documents, mixed layouts, charts that need interpretation, two-minute video clips, agentic tool use. The 1M-token context and 12B active parameters are the reason it exists, and no on-device model substitutes for them.

如果你拿不准，起决定作用的问题不是准确率，而是图像是否被允许离开设备。这个答案会在参考任何基准测试之前就决定选哪个模型。

MiniCPM-V 4.7 35B-A3B 的模型卡是这些端侧宣称的来源——每秒 13+ token、约 3 GB 的占用、day-0 芯片列表以及量化格式——所有这些都是该模型自行报告的结果。在决定采用之前先阅读芯片列表，因为其中的“支持”意味着存在一条运行时路径，而不是你的特定板卡已经过基准测试。

N​VIDIA's N​emotron developer page is where the 4× H100 / 2× GB200 guidance and the runtime versions are published, along with the link to the technical report behind the benchmark table. The version numbers in the card are exact for a reason: vLLM v0.32.0 and SGLang v0.5.20 with EAGLE topk 1 are the validated combination, and drifting from them is how a working deployment becomes a debugging session.

The larger point is that the vision-language field has stopped converging on one shape. OpenBMB is pushing a 3B-active model into pockets and N​VIDIA is pushing a 12B-active model across a million tokens of context, in the same month, under licences that both permit commercial use. Picking between them is picking a deployment, and the deployment is decided by where the pixels live.

基准测试会过时，价格会变动，但失败记录不会：这一类模型背后有据可查的 AI 事故与智能体故障，均收录于OrcaRouter AI 事故档案库，其中每条记录都带有各自的来源、严重程度、置信度评级和争议标记。