NVIDIA
Operating mode
Use this to bias workload recommendations toward responsiveness, background autonomy, lighter serving, or multi-GPU scale-out.
Current mode
Balanced
Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.
NVIDIA B100 是 Blackwell 数据中心 GPU，设计为现有 HGX H100 基础设施的直接升级，配备 192 GB HBM3e，带宽 8000 GB/s，提供 1750 TFLOPS FP16。作为 700W 低功耗 Blackwell 变体，可在与现有 H100 SXM 机架相同的散热包络内运行，同时提供更多 VRAM 和更高算力。
Beyond LLMs
What AI tasks this GPU can handle — from text generation to image and video creation.
| Capability | Status | Representative Model |
|---|---|---|
| LLM Chat (7B) | Runs natively | Llama 3.1 8B Q4 |
| LLM Coding (30B) | Runs natively | Qwen 3 30B Q4 |
| LLM Large (70B) | Runs natively | Llama 3.1 70B Q4 |
| Image Gen (SDXL) | Runs natively | SDXL 1.0 FP16 |
| Image Gen (Flux) | Runs natively | Flux.1 Dev FP16 |
| Image Gen (SD 3.5) | Runs natively | SD 3.5 Large FP16 |
| Video Short (25f) | Runs natively | LTX Video 2B |
| Video Long (100f) | Runs natively | Wan Video 14B |
Architecture
Blackwell is NVIDIA's fifth-generation RTX architecture, built on TSMC's 4NP process. It introduces 5th-generation Tensor Cores with native FP4 precision support, enabling double the inference throughput per watt compared to Ada Lovelace's FP8 operations. Key innovations include the Neural Rendering Pipeline for AI-driven shading and the debut of GDDR7 memory in consumer GPUs.
AI Relevance
FP4 Tensor Cores deliver the highest tokens-per-watt efficiency in any consumer architecture. Native FP4 quantization means models can run at lower precision with minimal quality loss, effectively doubling the effective VRAM for model weights.
购买建议
本地 AI 的绝佳选择
能良好运行 50 个顶级模型中的 40 个 — 本地推理的全能之选。
192.0 GB
VRAM
$35,000
建议零售价
$182/GB
每 GB VRAM 成本
最适合此 GPU 的模型
What will limit you first
This setup is broadly balanced for this model.
No major red flags
This recommendation has enough memory headroom and acceptable estimated speed for the selected workload.
Best upgrade itinerary
Unlocks 4 additional models that do not fit on the current setup.
想要更多余量？ AMD Instinct MI325X 256GB (256.0 GB VRAM) 是下一步升级选择。
Chat
SQwen 3.5 122B A10B matches Chat and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, lm-studio.
Coding
SDevstral 2 123B Instruct is a specialized fit for Coding. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, lm-studio.
Agentic Coding
SDevstral 2 123B Instruct is a specialized fit for Agentic Coding. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, lm-studio.
Reasoning
SDevstral 2 123B Instruct matches Reasoning and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, lm-studio.
RAG
SQwen 3.5 122B A10B matches RAG and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, lm-studio.
Devstral 2 123B Instruct
DeepSeek V4 Flash
Qwen 3.5 122B A10BMistral Small 4 119B
GPT-OSS 120BPixtral Large 124B
Command A 111BQwen 3 235B A22BQwen 2.5 VL 72BQwen3-Coder-NextQwen3-Coder 30B A3B InstructQwen 3.6 35B A3BLeanstral 119B A6BQwen 3.5 27BQwen3-VL 30B A3B InstructQwen 3.6 27BQwen 3.5 35B A3BQwen 3 32BMagistral Small 2507Devstral Small 2 24B InstructQwen 3 30B A3BQwen 3.5 9BNemotron 3 Nano 30BQwen 3 14BDevstral Small 1.1
Phi-4-reasoning-plus 14BQwen 3 8B
Gemma 4 31BGPT-OSS 20BNemotron Cascade 2 30B A3BQwen 3.5 4B
EXAONE 4.0 32BGemma 4 26B A4BMinistral 3 14BNemotron Nano 8BPhi-4 Mini Reasoning 4BDeepSeek Coder V2 236B
Jina Embeddings v3
BGE M3Qwen 3.5 397B A17B
Kimi K2.5Kimi K2.6DeepSeek V4 Pro
GLM-5.1GLM-5DeepSeek V3.2Qwen3-Coder 480B A35B InstructDeepSeek R1 671BDeepSeek V3.1 671B触手可及
高质量模型，只需稍多一点内存
Image & Video Generation
52 of 52 models can generate images or video on your B100 192GB
| Model | Max Resolution | Gen Time | Grade |
|---|---|---|---|
| SD TurboImage | 512×512 | 0ms | S |
| Stable Diffusion 1.5Image | 512×768 | 0ms | S |
| Realistic Vision v5.1Image | 512×768 | 0ms | S |
| DreamShaper 8Image | 512×768 | 0ms | S |
| LCM DreamShaper v7Image | 512×768 | 0ms | S |
| PixArt-SigmaImage | 1024×1024 | 200ms | S |
| FramePack I2VVideo | 1280×720 | 300ms/frame | S |
| SDXL TurboImage | 512×512 | 0ms | S |
| SDXL LightningImage | 1024×1024 | 100ms | S |
| Stable Diffusion XL 1.0Image | 1024×1024 | 200ms | S |
| Playground v2.5Image | 1024×1024 | 300ms | S |
| RealVisXL v5.0Image | 1024×1024 | 200ms | S |
| DreamShaper XLImage | 1024×1024 | 200ms | S |
| Juggernaut XL v9Image | 1024×1024 | 200ms | S |
| Animagine XL 3.1Image | 1024×1024 | 200ms | S |
| Pony Diffusion V6 XLImage | 1024×1024 | 200ms | S |
| Animagine XL 4.0Image | 1024×1024 | 200ms | S |
| Illustrious XLImage | 1024×1024 | 200ms | S |
| Wan Video 2.1 1.3BVideo | 480×832 | 100ms/frame | S |
| Stable Diffusion 3.5 MediumImage | 1024×1024 | 300ms | S |
| Flux.2 Klein 4BImage | 1024×1024 | 100ms | S |
| LTX Video 2BVideo | 1280×720 | 100ms/frame | S |
| KolorsImage | 1024×1024 | 300ms | S |
| Stable CascadeImage | 1024×1024 | 400ms | S |
| AuraFlow v0.3Image | 1536×1536 | 800ms | S |
| Stable Diffusion 3.5 LargeImage | 1024×1024 | 900ms | S |
| Stable Diffusion 3.5 Large TurboImage | 1024×1024 | 200ms | S |
| CogVideoX 2BVideo | 720×480 | 100ms/frame | S |
| HunyuanVideoVideo | 720×1280 | 300ms/frame | S |
| ChromaImage | 1024×1024 | 200ms | S |
| Z-Image TurboImage | 1536×1536 | 200ms | S |
| Flux.1 DevImage | 1024×1024 | 800ms | S |
| Flux.1 SchnellImage | 1024×1024 | 100ms | S |
| LTX Video 13BVideo | 1280×720 | 300ms/frame | S |
| Flux.1 Kontext DevImage | 1024×1024 | 800ms | S |
| AnimateDiff v1.5.3Video | 512×768 | 100ms/frame | S |
| Cosmos Diffusion 7BVideo | 1024×576 | 200ms/frame | S |
| CogVideoX 5BVideo | 720×480 | 200ms/frame | S |
| Wan2.2 TI2V 5BVideo | 832×480 | 200ms/frame | S |
| Flux.2 Klein 9BImage | 1024×1024 | 100ms | S |
| Flux.1 Fill DevImage | 1024×1024 | 700ms | S |
| Mochi 1 PreviewVideo | 848×480 | 300ms/frame | S |
| HunyuanVideo 1.5Video | 720×1280 | 300ms/frame | S |
| Helios 14BVideo | 1280×720 | 300ms/frame | S |
| SkyReels V2 14BVideo | 1280×720 | 300ms/frame | S |
| Wan Video 2.1 14BVideo | 720×1280 | 300ms/frame | S |
| Wan Video 2.2 14BVideo | 720×1280 | 300ms/frame | S |
| Qwen ImageImage | 1024×1024 | 300ms | S |
| Qwen Image EditImage | 1024×1024 | 300ms | S |
| Flux.2 DevImage | 1024×1024 | ~8s | S |
| MAGI-1Video | 1280×720 | 400ms/frame | S |
| HunyuanImage 3.0Image | 1024×1024 | 500ms | B |
Image models estimated at 1024×1024 (28 steps, FP16). Video models estimated at 768×512 (25 frames, 30 steps, FP16). Actual performance varies with runtime and system load.
Multi-GPU scaling
Scale out with multiple GPUs for larger models. NVLink provides 1800 GB/s inter-GPU bandwidth with 7% overhead.
| Config | Effective memory | Models that fit | Est. bandwidth |
|---|---|---|---|
| 1× B100 | 192 GB | 359/374 | 8,000 GB/s |
| 2× B100 | 384 GB | 366/374 | 14,880 GB/s |
| 4× B100 | 768 GB | 373/374 | 29,760 GB/s |
| 8× B100 | 1536 GB | 374/374 | 59,520 GB/s |
Model counts use default quantization at coding workload settings. Multi-GPU scaling factor: 0.93× per additional GPU.
Upgrade paths
See what you unlock with more powerful hardware
升级选项
8× B100 192GBMulti-GPUUnlocks 15 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 171%.
NVLink gives this scale-out path a cleaner inter-GPU story than PCIe-only builds.
~$35,000 MSRP
Unlocks 4 additional models that do not fit on the current setup.
~$20,000 MSRP
Unlocks 5 additional models that do not fit on the current setup.
~$8,000 MSRP
B100 192GB (192 GB VRAM) can run these top models: Devstral 2 123B Instruct (score: 96/100), DeepSeek V4 Flash (score: 96/100), Qwen 3.5 122B A10B (score: 95/100). See the full compatibility list above.
B100 192GB has 192 GB of VRAM available for AI model inference. This determines which models and quantization levels you can run locally.
Yes, B100 192GB is excellent for running LLMs locally with top compatibility scores above 80/100.
For coding on B100 192GB, we recommend Devstral 2 123B Instruct. It achieves 76.1 tokens per second with 212K context window. Devstral 2 123B Instruct is a specialized fit for Coding. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, lm-studio.
There are 3 upgrade path(s) from B100 192GB: B100 192GB, AMD Instinct MI325X 256GB. Upgrading would unlock larger models and faster inference speeds.
Yes, B100 192GB with 192 GB of usable memory can run Flux.1 Dev at FP16 natively. Flux is a 12B parameter diffusion transformer that produces high-quality images. You can also run the Schnell variant for faster generation.
B100 192GB (192 GB VRAM) can handle various AI generation tasks beyond LLMs. For image generation, SDXL and Stable Diffusion 3.5 run well. Flux.1 Dev also runs natively for state-of-the-art image quality. For video, LTX Video 2.3 can generate short clips. Check the AI Capability Matrix above for detailed compatibility.
B100 192GB is excellent for AI image generation. With 192 GB of usable memory, it runs all major diffusion models including Flux.1, SDXL, and Stable Diffusion 3.5 at full precision. You can generate high-resolution images quickly and even handle video generation models.
Yes, B100 192GB with 192 GB of usable memory can run Qwen 3.5 27B at Q8 (near-lossless, ~28.9 GB) or even FP16 (~55.4 GB) depending on your context needs. This setup provides an excellent experience with this model. Use Ollama or vLLM for best results.
With 192 GB VRAM on B100 192GB, use Q8_0 for most models — it is near-lossless and you have the memory for it. For 70B+ models, Q6_K offers excellent quality. Reserve Q4_K_M for 100B+ models or when you need maximum context length.
B100 192GB already has strong memory bandwidth, so the next limit is often memory capacity and context headroom rather than raw decode speed. For local LLMs, fit first and bandwidth second is the right mental model.
B100 192GB supports up to 8× GPU scaling via NVLink at 1800 GB/s. With 8× GPUs, you get 1536 GB effective memory with a 0.93× scaling factor per GPU. This enables running models like Qwen 3.5 397B A17B and Kimi K2.5 that don't fit on a single card.
NVLink is recommended for B100 192GB multi-GPU inference, providing 1800 GB/s interconnect bandwidth with only 7% scaling overhead. PCIe-only setups work but have higher overhead (~25%) due to limited inter-GPU bandwidth.
Compare with similar
