@meta
  v: 1
  route: /gpu/rtx-3070
  generated: 2026-06-10T09:30:33.281Z

@intent
  purpose:    Show what AI models fit on the RTX 3070.
  audience:   ai-engineer, self-hoster, homelab-buyer
  capability: inspect_specs, find_fitting_models, compare_to_other_gpus

@state
  slug: rtx-3070
  name: RTX 3070
  vram_gb: 8
  bandwidth_gbs: 448
  fp16_tflops: 81
  budget_8k_gb: 5.1
  category: rtx-30
  summary: "8GB ceiling. Sticks to 7B at Q5 or 13B at Q3. VRAM is the bottleneck, not silicon."
  top_fitting_models[5]: qwen-3-5-9b, gemma-2-9b, llama-3-1-8b, granite-8b, mistral-7b

@actions
  - id: compare
    method: GET
    href: /gpu/rtx-4060
  - id: compare
    method: GET
    href: /gpu/rtx-3080
  - id: compare
    method: GET
    href: /gpu/arc-b580
  - id: compare
    method: GET
    href: /gpu/rtx-3060
  - id: view_calculator
    method: GET
    href: /#calculator
  - id: view_math
    method: GET
    href: /the-math

@context
  > 8GB ceiling. Sticks to 7B at Q5 or 13B at Q3. VRAM is the bottleneck, not silicon.

@nav
  self:      /gpu/rtx-3070
  parents:   [/]
  peers:     [/gpu/rtx-4060, /gpu/rtx-3080, /gpu/arc-b580, /gpu/rtx-3060]
  drilldown: /the-math