Skip to main content

AI Compute Card Full Comparison Table (100+ Models)

Data continuously updated. Spot an error? Submit an Issue.

Quick Filter​

ScenarioRecommended Models
Trillion-parameter training (GPT-4 class)NVIDIA Rubin R200, B300 Ultra, AMD MI400, Google TPU Ironwood
10B–100B parameter trainingNVIDIA H100, H200, B200, AMD MI300X, MI325X
China market (domestic alternatives)Huawei Ascend 950, 910C, 920, Cambricon MLU690
High-throughput inferenceNVIDIA L40S, L4, H200 (inference mode), Intel Crescent Island
Edge AINVIDIA Jetson Orin, Google Edge TPU, Hailo-8L

Datacenter Training GPU​

ModelFP8 ComputeFP16 ComputeMemoryMemory BandwidthTDPRelease Date
NVIDIA Rubin R20050 PFLOPS (FP4)~10 PFLOPS288GB HBM422 TB/s~1,800W2026 H2
NVIDIA B300 Ultra14 PFLOPS~7 PFLOPS288GB HBM3e8 TB/s1,400W2025 Q4
NVIDIA B2009 PFLOPS~4.5 PFLOPS192GB HBM3e8 TB/s1,000W2025 Q2
NVIDIA B1007 PFLOPS~3.5 PFLOPS192GB HBM3e8 TB/s700W2024 Q4
NVIDIA H2003,958 TFLOPS1,979 TFLOPS141GB HBM3e4.8 TB/s700W2024 Q2
NVIDIA H100 SXM53,958 TFLOPS1,979 TFLOPS80GB HBM33.35 TB/s700W2022 Q3
AMD MI40040 PFLOPS (FP4)~10 PFLOPS432GB HBM419.6 TB/s~1,000W2026 H2
AMD MI355X10.1 PFLOPS (MXFP6)~5 PFLOPS288GB HBM3e8 TB/s1,400W2025 H2
AMD MI350X9.2 PFLOPS (MXFP6)~4.6 PFLOPS288GB HBM3e8 TB/s750W2025 H2
AMD MI325X2,614 TFLOPS1,307 TFLOPS256GB HBM3e6.48 TB/s750W2024 Q4
AMD MI300X2,614 TFLOPS1,307 TFLOPS192GB HBM35.3 TB/s750W2023 Q4
Huawei Ascend 950PR1 PFLOPS (FP8)~500 TFLOPS128GB HiBL (in-house)~3 TB/s~400W2026 H1
Huawei Ascend 950DT1 PFLOPS (FP8)~500 TFLOPS144GB HiZQ (in-house)4 TB/s~500W2026 H1
Huawei Ascend 920900+ TFLOPS (BF16)~450 TFLOPS~96GB HBM4 Tbps~400W2025 H2
Huawei Ascend 910C780 TFLOPS (BF16)~390 TFLOPS128GB HBM2e (dual-die)~1.2 TB/s~310W2025 H1
Huawei Ascend 910B320 TFLOPS (FP16)320 TFLOPS64GB HBM2e1.2 TB/s310W2023

Datacenter Inference GPU​

ModelFP8 ComputeINT8 ComputeMemoryTDPUse Case
NVIDIA L40S733 TFLOPS1,466 TOPS48GB GDDR6350WDatacenter inference
NVIDIA L4242 TFLOPS485 TOPS24GB GDDR672WEdge inference
NVIDIA L2~203 TFLOPS~406 TOPS16GB GDDR675WLow-power inference
NVIDIA RTX 6000 Ada1,452 TFLOPS2,905 TOPS48GB GDDR6300WWorkstation inference
NVIDIA T465 TFLOPS130 TOPS16GB GDDR670WEntry-level inference

AI Training ASIC (TPU / Gaudi / Trainium)​

ModelVendorCompute (BF16)MemoryInterconnect BandwidthRelease Date
Google TPU v6e (Trillium)Google918 TFLOPS32GB HBM1.6 Tb/s2024
Google TPU Ironwood (v7)Google~2,000 TFLOPS192GB HBM~5 Tb/s2026 H1
Google TPU 8t (Training)Google~1,200 TFLOPS64GB+ HBM~3 Tb/s2026 H1
Google TPU 8i (Inference)Google~1,500 TOPS64GB+ HBM~3 Tb/s2026 H1
Intel Gaudi 3Intel1,600 TFLOPS128GB SRAM2.4 Tb/s2024 Q2
Intel Crescent IslandIntelTBD480GB LPDDR5xTBD2026 H2
AWS Trainium 3AWS~5.7 PFLOPS~144GB~4.5 Tb/s2025 Q4
AWS Trainium 2AWS~1,000 TFLOPS64GB~1.6 Tb/s2024

Wafer-Scale Training​

ModelVendorTransistorsOn-Chip MemoryFP16 ComputeRelease Date
Cerebras WSE-4Cerebras4 trillion44GB SRAM125 PFLOPS2026
Cerebras WSE-3Cerebras4 trillion40GB SRAM125 PFLOPS2024
Cerebras WSE-2Cerebras2.6 trillion40GB SRAM85 PFLOPS2022

Edge AI & On-Device NPU​

ModelVendorCompute (TOPS)PowerUse Case
NVIDIA Jetson ThorNVIDIA2,070 TOPS130WRobotics / autonomous driving
NVIDIA Jetson Orin AGXNVIDIA275 TOPS60WEdge inference
Google Edge TPU (Dev Board)Google4 TOPS2WIoT on-device inference
Hailo-8LHailo13 TOPS1.5WOn-device vision AI
Qualcomm AI 100Qualcomm70 TOPS15WDatacenter edge inference
Huawei Ascend 310Huawei22 TOPS8WOn-device inference

Innovative Architectures​

ModelArchitecture TypeKey FeatureVendor
Groq LPU (LPU v2)LPU (Language Processing Unit)Ultra-low latency inference (~500 tok/s)Groq
Graphcore IPU (Bow POD)IPU (Intelligence Processing Unit)Native graph computing, 1,400 IPU coresGraphcore
Tesla Dojo (D1)Distributed training waferIntegrated auto-labeling + model trainingTesla
Apple M5 Ultra (Neural Engine)SoC + NPUOn-device 50 TOPS, unified memoryApple
Akida2 (AKD1000)Spiking Neural Network (SNN)Ultra-low-power neuromorphicBrainChip

Pricing Reference (Cloud On-Demand)​

ModelOn-Demand Price (USD/hr)Reserved Price (USD/hr)Purchase Price (USD)
NVIDIA B200$8.87~$5.50~$40,000
NVIDIA H200$5.87~$3.80~$30,000
NVIDIA H100$4.20~$2.80~$25,000
AMD MI300X——~$15,000
Huawei Ascend 910C——Domestic pricing

Note: Prices fluctuate with market supply and demand. Purchase prices are affected by export controls. The above data is for reference only.

Purchasing Guide​

By Model Scale​

  • Trillion-parameter (GPT-4 class): NVIDIA B300 Ultra / Rubin R200, AMD MI400 (2026 H2)
  • 10B–100B parameters (Llama 70B, Qwen 72B): NVIDIA H100 / H200, AMD MI300X / MI325X
  • 1B–10B parameters (Llama 7B–13B): NVIDIA H100, A100 80GB
  • Small models / inference: NVIDIA L40S, L4, T4

By Region​

  • North America / Europe: NVIDIA + AMD, freely available
  • China: Huawei Ascend 950 / 910C / 920 / Cambricon MLU690 (domestic alternatives)
  • Cloud (no hardware preference): Any vendor, choose by price

← Back to Home | Roadmap →