@meta
  v: 1
  route: /gpu/rtx-4090
  generated: 2026-06-10T07:49:26.276Z

@intent
  purpose:    Show what AI models fit on the RTX 4090.
  audience:   ai-engineer, self-hoster, homelab-buyer
  capability: inspect_specs, find_fitting_models, compare_to_other_gpus

@state
  slug: rtx-4090
  name: RTX 4090
  vram_gb: 24
  bandwidth_gbs: 1008
  fp16_tflops: 330
  budget_8k_gb: 18
  category: rtx-40
  summary: "The consumer flagship. Comfortably runs Llama-3.1 70B at Q4_K_M, Mixtral 8x7B at Q5, and any 30B-class model at Q8_0."
  top_fitting_models[5]: qwen-3-6-35b-a3b, yi-34b, qwen-2-5-32b, qwen-2-5-coder-32b, qwen-3-30b-a3b

@actions
  - id: compare
    method: GET
    href: /gpu/arc-pro-b60
  - id: compare
    method: GET
    href: /gpu/rtx-3090
  - id: compare
    method: GET
    href: /gpu/rtx-3090-ti
  - id: compare
    method: GET
    href: /gpu/rx-7900-xtx
  - id: view_calculator
    method: GET
    href: /#calculator
  - id: view_math
    method: GET
    href: /the-math

@context
  > The consumer flagship. Comfortably runs Llama-3.1 70B at Q4_K_M, Mixtral 8x7B at Q5, and any 30B-class model at Q8_0.

@nav
  self:      /gpu/rtx-4090
  parents:   [/]
  peers:     [/gpu/arc-pro-b60, /gpu/rtx-3090, /gpu/rtx-3090-ti, /gpu/rx-7900-xtx]
  drilldown: /the-math