99↗tokens.

Instinct MI350X OAM

Instinct MI350X OAM, server card/module, 288 GB HBM3e ECC

Physical memory288GiB
Available in calculator 286.0GiB
Peak bandwidth 8000GB/s
Use in calculator

Requires a compatible server, cooling and power delivery. Capacity is for one full GPU; a MIG/vGPU partition exposes less. Runtime support and actual free memory must be checked.

Specifications and setup

Memory
288 GiB HBM3e ECC
Connection
OAM module; Infinity Fabric on compatible server baseboard
Runtime
ROCm/HIP support depends on the exact GPU, OS and runtime release; Vulkan is another backend where supported.
Power
1000 W. Maximum/reference board power; excludes the host system and is not measured inference power
Bandwidth basis
Manufacturer peak memory bandwidth; not a measured sustained rate or a token/s estimate.
System reserve
Planner assumption: 2 GiB per device for driver/display and allocation headroom. Runtime buffers are modeled separately; actual free memory must be checked.

Model fit at Q4 and 8K

One request, FP16 conversation cache, default reserves. “Not verified” means the model’s cache formula needs more research.

ModelMemory neededAvailableEstimateOpen plan
Qwen3.5 0.8B0.9B · Dense1.6 GiB286.0 GiBRoom to spare
Qwen3.5 2B2.3B · Dense2.4 GiB286.0 GiBRoom to spare
SmolLM3 3B3.1B · Dense3.3 GiB286.0 GiBRoom to spare
Granite 4.2 3B3.7B · Dense3.7 GiB286.0 GiBRoom to spare
Ministral 3 3B3.8B · Dense4.0 GiB286.0 GiBRoom to spare
Qwen3.5 4B4.7B · Dense3.9 GiB286.0 GiBRoom to spare
Gemma 4 E2B5.1B · Dense3.9 GiB286.0 GiBConditional · Room to spare
OLMo 3 7B Instruct7.3B · Dense7.6 GiB286.0 GiBConditional · Room to spare
Gemma 4 E4B8.0B · Dense5.6 GiB286.0 GiBConditional · Room to spare
Granite 4.2 8B8.8B · Dense7.2 GiB286.0 GiBRoom to spare
Nemotron Nano 9B v28.9B · Dense6.2 GiB286.0 GiBRoom to spare
Ministral 3 8B8.9B · Dense7.0 GiB286.0 GiBRoom to spare
Qwen3.5 9B9.7B · Dense6.7 GiB286.0 GiBRoom to spare
GLM-4.6V Flash 9B10.3B · Dense7.1 GiB286.0 GiBRoom to spare
Gemma 4 12B12.0B · Dense8.1 GiB286.0 GiBConditional · Room to spare
Ministral 3 14B13.9B · Dense10.0 GiB286.0 GiBRoom to spare
Phi-4 Reasoning Plus 14B14.7B · Dense10.8 GiB286.0 GiBRoom to spare
gpt-oss 20B20.9B · Mixture of experts12.9 GiB286.0 GiBConditional · Room to spare
ERNIE 4.5 21B-A3B Thinking21.8B · Mixture of experts13.6 GiB286.0 GiBRoom to spare
Devstral Small 2 24B24.0B · Dense15.7 GiB286.0 GiBRoom to spare
Gemma 4 26B-A4B25.8B · Mixture of experts15.9 GiB286.0 GiBConditional · Room to spare
Qwen3.8 27B27.8B · Dense17.4 GiB286.0 GiBRoom to spare
Granite 4.2 30B29.3B · Dense19.7 GiB286.0 GiBRoom to spare
Qwen3 Coder 30B-A3B30.5B · Mixture of experts19.2 GiB286.0 GiBRoom to spare
GLM-4.7 Flash 30B-A3B31.2B · Mixture of experts19.3 GiB286.0 GiBConditional · Room to spare
Gemma 4 31B31.3B · Dense20.3 GiB286.0 GiBConditional · Room to spare
Nemotron 3 Nano 30B-A3B31.6B · Mixture of experts19.2 GiB286.0 GiBRoom to spare
OLMo 3.1 32B Think32.2B · Dense20.7 GiB286.0 GiBConditional · Room to spare
Qwen3.6 35B-A3B36.0B · Mixture of experts21.9 GiB286.0 GiBRoom to spare
Kimi Linear 48B-A3B49.1B · Mixture of experts30.8 GiB286.0 GiBConditional · Room to spare
Llama 3.3 70B70.6B · Dense45.1 GiB286.0 GiBRoom to spare
Qwen3 Coder Next 80B-A3B79.7B · Mixture of experts48.3 GiB286.0 GiBRoom to spare
Llama 4 Scout 109B-A17B108.6B · Mixture of experts67.1 GiB286.0 GiBConditional · Room to spare
gpt-oss 120B116.8B · Mixture of experts70.8 GiB286.0 GiBConditional · Room to spare
Mistral Small 4 119B-A6.5B119.4B · Mixture of experts72.2 GiB286.0 GiBConditional · Room to spare
Nemotron 3 Super 120B-A12B123.6B · Mixture of experts74.8 GiB286.0 GiBRoom to spare
Devstral 2 123B125.0B · Dense78.2 GiB286.0 GiBRoom to spare
Qwen3.5 122B-A10B125.1B · Mixture of experts75.8 GiB286.0 GiBRoom to spare
Mistral Medium 3.5 128B127.7B · Dense79.8 GiB286.0 GiBRoom to spare
Qwen3.8 Flash Next180.0B · Mixture of experts181.0 GiB286.0 GiBConditional · Room to spare
MiniMax M2.7228.7B · Mixture of experts140.0 GiB286.0 GiBRoom to spare
DeepSeek V4 Flash 0731304.2B · Mixture of experts183.6 GiB286.0 GiBConditional · Room to spare
DeepSeek V4 Flash Vision304.6B · Mixture of experts183.9 GiB286.0 GiBConditional · Room to spare
GLM-5.3 Flash321.3B · Mixture of experts194.1 GiB286.0 GiBConditional · Room to spare
Qwen3.5 397B-A17B403.4B · Mixture of experts243.9 GiB286.0 GiBRoom to spare
MiniMax M3427.0B · Mixture of experts258.8 GiB286.0 GiBConditional · Tight fit
Nemotron 3 Ultra 550B-A55B560.5B · Mixture of experts338.8 GiB286.0 GiBOver capacity
Mistral Large 3 675B-A41B675.0B · Mixture of experts407.9 GiB286.0 GiBConditional · Over capacity
GLM-5.3753.3B · Mixture of experts455.3 GiB286.0 GiBConditional · Over capacity
Kimi K2.7 Code1,026.9B · Mixture of experts620.3 GiB286.0 GiBConditional · Over capacity
DeepSeek V4 Pro 08131,650.5B · Mixture of experts996.2 GiB286.0 GiBConditional · Over capacity
Qwen3.8 2.4T-A95B2,446.2B · Mixture of experts1,477.5 GiB286.0 GiBOver capacity
Kimi K32,779.9B · Mixture of experts1,678.3 GiB286.0 GiBConditional · Over capacity