@meta
  v: 1
  route: /model/gemma-4-26b
  generated: 2026-06-10T08:49:56.802Z

@intent
  purpose:    Show what hardware can run Gemma 4 26B A4B.
  audience:   ai-engineer, self-hoster, model-evaluator
  capability: inspect_model_specs, find_compatible_gpus, compare_similar_models, open_huggingface

@state
  slug: gemma-4-26b
  name: Gemma 4 26B A4B
  hf_repo: google/gemma-4-26B-A4B-it
  params_b: 26
  active_params_b: 4
  family: Google
  type: moe
  context_k: 128
  fp16_gb: 52
  summary: "Google's 2026 MoE: 26B total, 4B active. Vision-ready."
  smallest_fitting_gpu_q4: rtx-3090
  smallest_fitting_gpu_fp16: h100

@actions
  - id: open_huggingface
    method: GET
    href: https://huggingface.co/google/gemma-4-26B-A4B-it
  - id: view_compatible_gpus
    method: GET
    href: /model/gemma-4-26b
  - id: compare_similar
    method: GET
    href: /model/qwen-3-6-27b
  - id: view_calculator
    method: GET
    href: /#calculator

@context
  > Google's 2026 MoE: 26B total, 4B active. Vision-ready.

@nav
  self:      /model/gemma-4-26b
  parents:   [/]
  peers:     [/model/qwen-3-6-27b, /model/mistral-small-3, /model/qwen-3-30b-a3b, /model/gpt-oss-20b]
  drilldown: /the-math