BBrainOutput
NVIDIA·PlatformProvisional

NVIDIA GH200 (Grace Hopper): Specs & Local-AI Compatibility

Grace-Hopper superchip with large coherent memory for big-model serving.

Some details here are provisional (placeholder). Treat specs as approximate and verify against the manufacturer before relying on them or purchasing.

Specs

Memory
480 GB unified
Memory type
HBM3e + LPDDR5X
Bandwidth
to verify
Approx FP16
990 TFLOPS
Architecture
Grace Hopper
Process
TSMC 4N
Power
1,000 W
Launch
2024

Models this chip can run

Open models graded for a single NVIDIA GH200 (Grace Hopper), best fit first.

  • Gemma 2 27B
    Gemma · ~27B · 8K ctx · Gemma Terms of Use

    Fits at FP16 (~54GB) with ~282GB headroom — about 6 concurrent instances.

    FP16 · ~54GBRuns well
  • Gemma 3 27B
    Gemma 3 · ~27B · 128K ctx · Gemma Terms of Use

    Fits at FP16 (~54GB) with ~282GB headroom — about 6 concurrent instances.

    FP16 · ~54GBRuns well
  • Mistral Small 24B
    Mistral · ~24B · 32K ctx · Apache-2.0

    Fits at FP16 (~48GB) with ~288GB headroom — about 7 concurrent instances.

    FP16 · ~48GBRuns well
  • DeepSeek-Coder V2 (class)
    DeepSeek · ~16B · 128K ctx · DeepSeek License

    Fits at FP16 (~33GB) with ~303GB headroom — about 10 concurrent instances.

    FP16 · ~33GBRuns well
  • StarCoder2 15B
    StarCoder · ~15B · 16K ctx · BigCode OpenRAIL-M

    Fits at FP16 (~30GB) with ~306GB headroom — about 11 concurrent instances.

    FP16 · ~30GBRuns well
  • Qwen2.5 14B
    Qwen · ~14B · 128K ctx · Apache-2.0

    Fits at FP16 (~30GB) with ~306GB headroom — about 11 concurrent instances.

    FP16 · ~30GBRuns well
  • Qwen3 14B
    Qwen · ~14B · 128K ctx · Apache-2.0

    Fits at FP16 (~30GB) with ~306GB headroom — about 11 concurrent instances.

    FP16 · ~30GBRuns well
  • Phi-3 Medium (14B)
    Phi · ~14B · 128K ctx · MIT

    Fits at FP16 (~28GB) with ~308GB headroom — about 12 concurrent instances.

    FP16 · ~28GBRuns well

Build a private AI Business OS on NVIDIA GH200 (Grace Hopper)

Run your own AI agents on hardware you control — private by design, no per-seat data leaving your premises. BrainOutput helps you pick the right machine and turn it into a working AI Business OS.

Explore the AI Business OS