AMD
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.
The AMD Instinct MI250 128GB is a CDNA 2 OAM-format accelerator with 128 GB of HBM2e memory spread across two GPU chiplets. It delivers 3.2 TB/s of aggregate memory bandwidth and full ROCm support, making it one of the most memory-bandwidth-rich AI platforms available. The MI250 is the non-XL variant of the MI250X, with similar memory but slightly lower compute. It enables inference of 405B+ models at Q4 with sufficient system bandwidth.
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
CDNA 2 powers the Instinct MI210 and MI250/MI250X accelerators. It introduced multi-die packaging with up to 128 GB HBM2e and Infinity Fabric for die-to-die communication.
AI Relevance
With up to 128 GB HBM2e memory and strong ROCm support, CDNA 2 GPUs can host large language models. The MI250X was used in the Frontier exascale supercomputer and supports major AI frameworks.
Kaufberatung
Ausgezeichnete Wahl für lokale KI
Führt 36 von 50 Top-Modellen gut aus — ein starker Allrounder für lokale Inferenz.
128.0 GB
VRAM
$19,000
UVP
$148/GB
Kosten pro GB VRAM
Beste Modelle für diese GPU
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 2 additional models that do not fit on the current setup.
Mehr Spielraum gewünscht? NVIDIA H200 141GB (141.0 GB VRAM) ist die nächste Stufe.
Chat
SThis model is a direct match for chat. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.
Coding
SThis model is a direct match for coding. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.
Agentic Coding
SThis model is still usable for agentic-coding, but it is not the most specialized pick. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.
Reasoning
SThis model is a direct match for reasoning. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.
RAG
SThis model is a direct match for rag. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.
Qwen 3.5 122B A10B
Mistral Small 4 119BDevstral 2 123B Instruct
GPT-OSS 120B
Command A 111BPixtral Large 124BLeanstral 119B A6BQwen3-Coder-NextQwen 2.5 VL 72BQwen 3.6 35B A3BQwen3-Coder 30B A3B InstructQwen 3.5 27BQwen3-VL 30B A3B InstructQwen 3.5 35B A3BQwen 3.6 27BQwen 3 32BMagistral Small 2507Devstral Small 2 24B InstructQwen 3 30B A3B
Nemotron 3 Nano 30BQwen 3.5 9BQwen 3 14BDevstral Small 1.1
Gemma 4 31B
Phi-4-reasoning-plus 14BNemotron Cascade 2 30B A3BQwen 3 8BGPT-OSS 20BQwen 3.5 4B
EXAONE 4.0 32BGemma 4 26B A4BMinistral 3 14BNemotron Nano 8BPhi-4 Mini Reasoning 4B
Jina Embeddings v3
BGE M3Qwen 3.5 397B A17B
Kimi K2.5Kimi K2.6
DeepSeek V4 ProDeepSeek V4 Flash
GLM-5.1GLM-5DeepSeek V3.2Qwen 3 235B A22BQwen3-Coder 480B A35B InstructDeepSeek Coder V2 236BDeepSeek R1 671BDeepSeek V3.1 671BFast erreichbar
Hochwertige Modelle, die etwas mehr Speicher benötigen
Image & Video Generation
52 of 52 models can generate images or video on your AMD Instinct MI250 128GB
| Model | Max Resolution | Gen Time | Grade |
|---|---|---|---|
| SD TurboImage | 512×512 | 100ms | S |
| Stable Diffusion 1.5Image | 512×768 | 300ms | S |
| Realistic Vision v5.1Image | 512×768 | 300ms | S |
| DreamShaper 8Image | 512×768 | 300ms | S |
| LCM DreamShaper v7Image | 512×768 | 100ms | S |
| PixArt-SigmaImage | 1024×1024 | ~1s | S |
| FramePack I2VVideo | 1280×720 | ~1.9s/frame | S |
| SDXL TurboImage | 512×512 | 100ms | S |
| SDXL LightningImage | 1024×1024 | 400ms | S |
| Stable Diffusion XL 1.0Image | 1024×1024 | ~1s | S |
| Playground v2.5Image | 1024×1024 | ~1.5s | S |
| RealVisXL v5.0Image | 1024×1024 | ~1.1s | S |
| DreamShaper XLImage | 1024×1024 | ~1.1s | S |
| Juggernaut XL v9Image | 1024×1024 | ~1.1s | S |
| Animagine XL 3.1Image | 1024×1024 | ~1.1s | S |
| Pony Diffusion V6 XLImage | 1024×1024 | ~1.1s | S |
| Animagine XL 4.0Image | 1024×1024 | ~1.1s | S |
| Illustrious XLImage | 1024×1024 | ~1.1s | S |
| Wan Video 2.1 1.3BVideo | 480×832 | 700ms/frame | S |
| Stable Diffusion 3.5 MediumImage | 1024×1024 | ~1.8s | S |
| Flux.2 Klein 4BImage | 1024×1024 | 300ms | S |
| LTX Video 2BVideo | 1280×720 | 900ms/frame | S |
| KolorsImage | 1024×1024 | ~2s | S |
| Stable CascadeImage | 1024×1024 | ~2.5s | S |
| AuraFlow v0.3Image | 1536×1536 | ~4.6s | S |
| Stable Diffusion 3.5 LargeImage | 1024×1024 | ~5.6s | S |
| Stable Diffusion 3.5 Large TurboImage | 1024×1024 | ~1s | S |
| CogVideoX 2BVideo | 720×480 | 900ms/frame | S |
| HunyuanVideoVideo | 720×1280 | ~1.9s/frame | S |
| ChromaImage | 1024×1024 | ~1s | S |
| Z-Image TurboImage | 1536×1536 | ~1s | S |
| Flux.1 DevImage | 1024×1024 | ~4.6s | S |
| Flux.1 SchnellImage | 1024×1024 | 900ms | S |
| LTX Video 13BVideo | 1280×720 | ~1.9s/frame | S |
| Flux.1 Kontext DevImage | 1024×1024 | ~5.1s | S |
| AnimateDiff v1.5.3Video | 512×768 | 500ms/frame | S |
| Cosmos Diffusion 7BVideo | 1024×576 | ~1.5s/frame | S |
| CogVideoX 5BVideo | 720×480 | ~1.3s/frame | S |
| Wan2.2 TI2V 5BVideo | 832×480 | ~1.3s/frame | S |
| Flux.2 Klein 9BImage | 1024×1024 | 500ms | S |
| Flux.1 Fill DevImage | 1024×1024 | ~4.3s | S |
| Mochi 1 PreviewVideo | 848×480 | ~1.7s/frame | S |
| HunyuanVideo 1.5Video | 720×1280 | ~1.6s/frame | S |
| Helios 14BVideo | 1280×720 | ~1.9s/frame | S |
| SkyReels V2 14BVideo | 1280×720 | ~1.9s/frame | S |
| Wan Video 2.1 14BVideo | 720×1280 | ~1.9s/frame | S |
| Wan Video 2.2 14BVideo | 720×1280 | ~1.9s/frame | S |
| Qwen ImageImage | 1024×1024 | ~1.7s | S |
| Qwen Image EditImage | 1024×1024 | ~1.7s | S |
| Flux.2 DevImage | 1024×1024 | ~47.9s | S |
| MAGI-1Video | 1280×720 | ~2.4s/frame | S |
| HunyuanImage 3.0Image | 256×256 | ~3s | D |
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.
Upgrade paths
See what you unlock with more powerful hardware
Upgrade-Optionen
NVIDIA H200 141GBNächste StufeUnlocks 2 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 29%.
ca. $30,000 MSRP
NVIDIA B200 180GBGrößter SprungUnlocks 8 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 57%.
ca. $30,000 MSRP
Unlocks 12 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 31%.
ca. $20,000 MSRP
Unlocks 13 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 47%.
ca. $8,000 MSRP
AMD Instinct MI250 128GB (128 GB VRAM) can run these top models: Qwen 3.5 122B A10B (score: 99/100), Mistral Small 4 119B (score: 97/100), Devstral 2 123B Instruct (score: 97/100). See the full compatibility list above.
AMD Instinct MI250 128GB has 128 GB of VRAM available for AI model inference. This determines which models and quantization levels you can run locally.
Yes, AMD Instinct MI250 128GB is excellent for running LLMs locally with top compatibility scores above 80/100.
For coding on AMD Instinct MI250 128GB, we recommend Qwen3-Coder-Next. It achieves 146.9 tokens per second with 256K context window. This model is a direct match for coding. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.
There are 4 upgrade path(s) from AMD Instinct MI250 128GB: NVIDIA H200 141GB, NVIDIA B200 180GB. Upgrading would unlock larger models and faster inference speeds.
Yes, AMD Instinct MI250 128GB with 128 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.
AMD Instinct MI250 128GB (128 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.
AMD Instinct MI250 128GB is excellent for AI image generation. With 128 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, AMD Instinct MI250 128GB with 128 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 128 GB VRAM on AMD Instinct MI250 128GB, 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.
AMD Instinct MI250 128GB 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.
Compare with similar
