@meta
  v: 1
  route: /model/gpt-oss-20b
  generated: 2026-06-10T08:50:23.793Z

@intent
  purpose:    Show what hardware can run gpt-oss 20B.
  audience:   ai-engineer, self-hoster, model-evaluator
  capability: inspect_model_specs, find_compatible_gpus, compare_similar_models, open_huggingface

@state
  slug: gpt-oss-20b
  name: gpt-oss 20B
  hf_repo: openai/gpt-oss-20b
  params_b: 20.9
  active_params_b: 3.6
  family: OpenAI
  type: moe
  context_k: 128
  fp16_gb: 41.8
  summary: OpenAI's August 2025 open-weight release. MoE; ships MXFP4-quantized at ~13GB on consumer cards.
  smallest_fitting_gpu_q4: rtx-4060-ti
  smallest_fitting_gpu_fp16: m2-max-64

@actions
  - id: open_huggingface
    method: GET
    href: https://huggingface.co/openai/gpt-oss-20b
  - id: view_compatible_gpus
    method: GET
    href: /model/gpt-oss-20b
  - id: compare_similar
    method: GET
    href: /model/mistral-small-3
  - id: view_calculator
    method: GET
    href: /#calculator

@context
  > OpenAI's August 2025 open-weight release. MoE; ships MXFP4-quantized at ~13GB on consumer cards.

@nav
  self:      /model/gpt-oss-20b
  parents:   [/]
  peers:     [/model/mistral-small-3, /model/gemma-4-26b, /model/starcoder2-15b, /model/qwen-3-6-27b]
  drilldown: /the-math