lingshu-core 0.10.0

# ─── Lingshu Model Catalog ─────────────────────────────────────────
#
# Single source of truth for all available models.
# Compiled into the binary as the default; user can override with
# ~/.lingshu/models.yaml (same format, merged on top).
#
# Kept at parity with edgecode's hermes_cli/models.py
#
# Format:
#   providers:
#     <provider_id>:
#       label: "Human Name"
#       default_model: "model-name"        # optional
#       models:
#         - model: "model-name"
#           context: 200000                 # tokens
#           tier: standard                  # fast | standard | reasoning
#           pricing:                        # optional, per million tokens
#             input: 3.0
#             output: 15.0
#             cache_read: 0.3
#             cache_write: 3.75

providers:

  # ─── Anthropic ────────────────────────────────────────────────────
  anthropic:
    label: "Anthropic"
    default_model: "claude-opus-4.6"
    models:
      - model: "claude-opus-4.6"
        context: 200000
        tier: reasoning
        pricing: { input: 15.0, output: 75.0, cache_read: 1.5, cache_write: 18.75 }
      - model: "claude-sonnet-4.6"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.3, cache_write: 3.75 }
      - model: "claude-3-5-haiku-latest"
        context: 200000
        tier: fast
        pricing: { input: 0.8, output: 4.0, cache_read: 0.08, cache_write: 1.0 }
      - model: "claude-3-5-sonnet-latest"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.3, cache_write: 3.75 }
      - model: "claude-3-opus-latest"
        context: 200000
        tier: reasoning
        pricing: { input: 15.0, output: 75.0, cache_read: 1.5, cache_write: 18.75 }
      - model: "claude-haiku-4.5"
        context: 200000
        tier: fast
        pricing: { input: 0.8, output: 4.0, cache_read: 0.08, cache_write: 1.0 }
      - model: "claude-haiku-4.5-20251001"
        context: 200000
        tier: fast
        pricing: { input: 0.8, output: 4.0, cache_read: 0.08, cache_write: 1.0 }
      - model: "claude-opus-4.5"
        context: 200000
        tier: reasoning
        pricing: { input: 15.0, output: 75.0, cache_read: 1.5, cache_write: 18.75 }
      - model: "claude-opus-4.5-20251101"
        context: 200000
        tier: reasoning
        pricing: { input: 15.0, output: 75.0, cache_read: 1.5, cache_write: 18.75 }
      - model: "claude-sonnet-4.5"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.3, cache_write: 3.75 }
      - model: "claude-sonnet-4.5-20250929"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.3, cache_write: 3.75 }

  # ─── Claude Pro (subscription OAuth) ─────────────────────────────
  # Use `lingshu auth add claude-pro` — routes to anthropic with OAuth token.
  claude-pro:
    label: "Claude Pro (OAuth)"
    default_model: "claude-sonnet-4.6"
    models:
      - model: "claude-opus-4.6"
        context: 200000
        tier: reasoning
      - model: "claude-sonnet-4.6"
        context: 200000
        tier: standard
      - model: "claude-sonnet-4.5"
        context: 200000
        tier: standard
      - model: "claude-haiku-4.5"
        context: 200000
        tier: fast

  # ─── GitHub Copilot ──────────────────────────────────────────────
  copilot:
    label: "GitHub Copilot"
    default_model: "auto"
    models:
      - model: "auto"
        context: 400000
        tier: standard
      - model: "claude-haiku-4.5"
        context: 200000
        tier: fast
      - model: "claude-opus-4.5"
        context: 200000
        tier: standard
      - model: "claude-opus-4.6"
        context: 200000
        tier: standard
      - model: "claude-opus-4.6-fast"
        context: 200000
        tier: standard
      - model: "claude-sonnet-4"
        context: 216000
        tier: standard
      - model: "claude-sonnet-4.5"
        context: 200000
        tier: standard
      - model: "claude-sonnet-4.6"
        context: 200000
        tier: standard
      - model: "gemini-2.5-pro"
        context: 128000
        tier: reasoning
      - model: "gemini-3-flash-preview"
        context: 128000
        tier: fast
      - model: "gemini-3.1-pro-preview"
        context: 128000
        tier: reasoning
      - model: "gpt-4.1"
        context: 128000
        tier: standard
      - model: "gpt-4o"
        context: 128000
        tier: standard
      - model: "gpt-5-mini"
        context: 264000
        tier: fast
      - model: "gpt-5.2"
        context: 264000
        tier: standard
      - model: "gpt-5.2-codex"
        context: 400000
        tier: standard
      - model: "gpt-5.3-codex"
        context: 400000
        tier: standard
      - model: "gpt-5.4"
        context: 400000
        tier: standard
      - model: "gpt-5.4-mini"
        context: 400000
        tier: fast
      - model: "raptor-mini"
        context: 264000
        tier: fast
      - model: "goldeneye"
        context: 264000
        tier: standard
      - model: "grok-code-fast-1"
        context: 128000
        tier: standard
      - model: "oswe-vscode-prime"
        context: 264000
        tier: standard

  # ─── ChatGPT Pro / Codex (subscription OAuth) ────────────────────
  chatgpt-pro:
    label: "ChatGPT Pro / Codex (OAuth)"
    default_model: "gpt-5.4"
    models:
      - model: "gpt-5.5"
        context: 400000
        tier: standard
      - model: "gpt-5.4"
        context: 400000
        tier: standard
      - model: "gpt-5.4-mini"
        context: 400000
        tier: fast
      - model: "gpt-5.3-codex"
        context: 400000
        tier: standard
      - model: "gpt-5.3-codex-spark"
        context: 400000
        tier: reasoning

  # ─── OpenAI ──────────────────────────────────────────────────────
  openai:
    label: "OpenAI"
    default_model: "gpt-5.4"
    models:
      - model: "gpt-3.5-turbo"
        context: 16000
        tier: fast
        pricing: { input: 0.50, output: 1.50 }
      - model: "gpt-4-turbo"
        context: 128000
        tier: standard
        pricing: { input: 10.0, output: 30.0 }
      - model: "gpt-4.1"
        context: 128000
        tier: standard
        pricing: { input: 2.0, output: 8.0, cache_read: 0.5 }
      - model: "gpt-4.1-mini"
        context: 128000
        tier: fast
        pricing: { input: 0.40, output: 1.60, cache_read: 0.10 }
      - model: "gpt-4.1-nano"
        context: 128000
        tier: fast
        pricing: { input: 0.10, output: 0.40, cache_read: 0.025 }
      - model: "gpt-4o"
        context: 128000
        tier: standard
        pricing: { input: 2.50, output: 10.0, cache_read: 1.25 }
      - model: "gpt-4o-mini"
        context: 128000
        tier: fast
        pricing: { input: 0.15, output: 0.60, cache_read: 0.075 }
      - model: "gpt-5"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.75 }
      - model: "gpt-5-codex"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.75 }
      - model: "gpt-5-mini"
        context: 200000
        tier: standard
        pricing: { input: 0.80, output: 3.20, cache_read: 0.20 }
      - model: "gpt-5-nano"
        context: 128000
        tier: fast
        pricing: { input: 0.15, output: 0.60, cache_read: 0.04 }
      - model: "gpt-5-pro"
        context: 200000
        tier: reasoning
        pricing: { input: 10.0, output: 40.0, cache_read: 2.50 }
      - model: "gpt-5.1"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.75 }
      - model: "gpt-5.1-codex"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.75 }
      - model: "gpt-5.1-codex-max"
        context: 256000
        tier: reasoning
        pricing: { input: 10.0, output: 40.0, cache_read: 2.5 }
      - model: "gpt-5.1-codex-mini"
        context: 200000
        tier: fast
        pricing: { input: 0.80, output: 3.20, cache_read: 0.20 }
      - model: "gpt-5.2"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.75 }
      - model: "gpt-5.2-codex"
        context: 200000
        tier: standard
        pricing: { input: 3.0, output: 15.0, cache_read: 0.75 }
      - model: "gpt-5.2-pro"
        context: 200000
        tier: reasoning
        pricing: { input: 10.0, output: 40.0, cache_read: 2.5 }
      - model: "gpt-5.4"
        context: 400000
        tier: standard
      - model: "gpt-5.4-mini"
        context: 400000
        tier: fast
      - model: "gpt-5.4-pro"
        context: 400000
        tier: reasoning
      - model: "gpt-5.4-nano"
        context: 400000
        tier: fast
      - model: "o1"
        context: 128000
        tier: reasoning
        pricing: { input: 15.0, output: 60.0, cache_read: 7.5 }
      - model: "o1-mini"
        context: 128000
        tier: reasoning
        pricing: { input: 3.0, output: 12.0, cache_read: 1.5 }
      - model: "o3"
        context: 200000
        tier: reasoning
        pricing: { input: 10.0, output: 40.0, cache_read: 2.5 }
      - model: "o3-mini"
        context: 128000
        tier: reasoning
        pricing: { input: 1.10, output: 4.40, cache_read: 0.55 }
      - model: "o3-pro"
        context: 200000
        tier: reasoning
        pricing: { input: 20.0, output: 80.0, cache_read: 5.0 }
      - model: "o4-mini"
        context: 200000
        tier: reasoning
        pricing: { input: 1.10, output: 4.40, cache_read: 0.55 }

  # ─── Google ──────────────────────────────────────────────────────
  # Uses Google AI Studio (GEMINI_API_KEY / GOOGLE_API_KEY).
  google:
    label: "Google"
    default_model: "gemini-2.5-flash"
    models:
      - model: "gemini-2.5-pro"
        context: 1000000
        tier: reasoning
        pricing: { input: 1.25, output: 10.0 }
      - model: "gemini-2.5-flash"
        context: 1000000
        tier: fast
        pricing: { input: 0.30, output: 2.50 }
      - model: "gemini-2.5-flash-lite"
        context: 1000000
        tier: fast
        pricing: { input: 0.10, output: 0.40 }
      - model: "gemini-2.0-flash"
        context: 1000000
        tier: fast
        pricing: { input: 0.15, output: 0.60 }
      - model: "gemini-2.0-flash-lite"
        context: 1000000
        tier: fast
        pricing: { input: 0.075, output: 0.30 }

  # ─── Hugging Face ────────────────────────────────────────────────
  huggingface:
    label: "Hugging Face"
    default_model: "Qwen/Qwen3.5-397B-A17B"
    models:
      - model: "Qwen/Qwen3.5-397B-A17B"
        context: 1000000
        tier: reasoning
      - model: "Qwen/Qwen3.5-35B-A3B"
        context: 1000000
        tier: standard
      - model: "deepseek-ai/DeepSeek-V3.2"
        context: 128000
        tier: standard
      - model: "moonshotai/Kimi-K2.5"
        context: 262000
        tier: reasoning
      - model: "MiniMaxAI/MiniMax-M2.5"
        context: 204000
        tier: standard
      - model: "zai-org/GLM-5"
        context: 128000
        tier: reasoning
      - model: "XiaomiMiMo/MiMo-V2-Flash"
        context: 256000
        tier: fast
      - model: "moonshotai/Kimi-K2-Thinking"
        context: 262000
        tier: reasoning

  # ─── Vertex AI ───────────────────────────────────────────────────
  # GA models are available on any region (e.g. us-central1).
  # Preview models marked (*global*) require GOOGLE_CLOUD_REGION=global;
  # Lingshu sets this automatically when the model name starts with "gemini-3".
  vertexai:
    label: "Vertex AI"
    default_model: "gemini-2.5-flash"
    models:
      # ── Gemini 2.x  (GA, regional + global) ──────────────────────
      - model: "gemini-2.5-pro"
        context: 1000000
        tier: reasoning
        pricing: { input: 1.25, output: 10.0 }
      - model: "gemini-2.5-flash"
        context: 1000000
        tier: fast
        pricing: { input: 0.30, output: 2.50 }
      - model: "gemini-2.5-flash-lite"
        context: 1000000
        tier: fast
        pricing: { input: 0.10, output: 0.40 }
      - model: "gemini-2.0-flash"
        context: 1000000
        tier: fast
        pricing: { input: 0.15, output: 0.60 }
      - model: "gemini-2.0-flash-lite"
        context: 1000000
        tier: fast
        pricing: { input: 0.075, output: 0.30 }
      # ── Gemini 3.x  (Preview, *global endpoint only*) ────────────
      # Requires GOOGLE_CLOUD_REGION=global (set automatically by Lingshu).
      - model: "gemini-3-flash-preview"
        context: 1048576
        tier: fast
        pricing: { input: 0.50, output: 3.0 }
      - model: "gemini-3.1-pro-preview"
        context: 1048576
        tier: reasoning
        pricing: { input: 2.0, output: 12.0 }
      - model: "gemini-3.1-flash-lite-preview"
        context: 1048576
        tier: fast
        pricing: { input: 0.25, output: 1.50 }

  # ─── AWS Bedrock ─────────────────────────────────────────────────
  # Uses the Bedrock runtime provider. Bare model IDs are resolved to the
  # correct cross-region inference profile when the backend requires it.
  bedrock:
    label: "AWS Bedrock"
    default_model: "amazon.nova-lite-v1:0"
    models:
      - model: "amazon.nova-lite-v1:0"
        context: 300000
        tier: fast
      - model: "amazon.nova-pro-v1:0"
        context: 300000
        tier: standard
      - model: "anthropic.claude-3-5-haiku-20241022-v1:0"
        context: 200000
        tier: fast
      - model: "anthropic.claude-3-5-sonnet-20241022-v2:0"
        context: 200000
        tier: standard
      - model: "anthropic.claude-3-7-sonnet-20250219-v1:0"
        context: 200000
        tier: reasoning
      - model: "anthropic.claude-4-sonnet-20250514-v1:0"
        context: 200000
        tier: standard
      - model: "anthropic.claude-4-opus-20250514-v1:0"
        context: 200000
        tier: reasoning
      - model: "meta.llama3-70b-instruct-v1:0"
        context: 128000
        tier: standard
      - model: "mistral.mistral-large-2407-v1:0"
        context: 32000
        tier: standard
      - model: "cohere.command-r-plus-v1:0"
        context: 128000
        tier: standard
      - model: "deepseek.r1-v1:0"
        context: 128000
        tier: reasoning
      - model: "qwen.qwen2-5-72b-instruct-v1:0"
        context: 131072
        tier: standard

  # ─── xAI ─────────────────────────────────────────────────────────
  xai:
    label: "xAI"
    default_model: "grok-4.20-beta"
    models:
      - model: "grok-2-vision-1212"
        context: 32000
        tier: standard
        pricing: { input: 2.0, output: 10.0 }
      - model: "grok-3"
        context: 131000
        tier: standard
        pricing: { input: 3.0, output: 15.0 }
      - model: "grok-3-mini"
        context: 131000
        tier: fast
        pricing: { input: 0.30, output: 0.50 }
      - model: "grok-4"
        context: 131000
        tier: standard
        pricing: { input: 3.0, output: 15.0 }
      - model: "grok-4-latest"
        context: 131000
        tier: standard
        pricing: { input: 3.0, output: 15.0 }
      - model: "grok-4-1-fast"
        context: 2000000
        tier: fast
        pricing: { input: 0.60, output: 2.40 }
      - model: "grok-code-fast-1"
        context: 131000
        tier: fast
        pricing: { input: 0.15, output: 0.60 }
      - model: "grok-4.20-beta"
        context: 2000000
        tier: reasoning

  # ─── SuperGrok (xAI OAuth subscription) ──────────────────────────
  # Uses `lingshu auth add grok` credentials from ~/.lingshu/auth.json.
  # This provider is intentionally separate from `xai` (API-key flow) so users
  # can explicitly choose subscription-backed routing in the model picker.
  super-grok:
    label: "SuperGrok (OAuth)"
    default_model: "grok-4.20-beta"
    models:
      - model: "grok-4.20-beta"
        context: 2000000
        tier: reasoning
      - model: "grok-4"
        context: 131000
        tier: standard
      - model: "grok-4-1-fast"
        context: 2000000
        tier: fast
      - model: "grok-code-fast-1"
        context: 131000
        tier: fast

  # ─── Z.AI / GLM ─────────────────────────────────────────────────
  zai:
    label: "Z.AI Platform"
    default_model: "glm-5"
    models:
      - model: "glm-5"
        context: 128000
        tier: reasoning
      - model: "glm-5-turbo"
        context: 128000
        tier: fast
      - model: "glm-4.5"
        context: 128000
        tier: reasoning
        pricing: { input: 0.50, output: 2.0 }
      - model: "glm-4.7"
        context: 128000
        tier: reasoning
        pricing: { input: 0.50, output: 2.0 }
      - model: "glm-4.7-flash"
        context: 128000
        tier: fast
        pricing: { input: 0.10, output: 0.40 }

  # ─── DeepSeek ────────────────────────────────────────────────────
  deepseek:
    label: "DeepSeek"
    default_model: "deepseek-chat"
    models:
      - model: "deepseek-v3.2"
        context: 128000
        tier: standard
      - model: "deepseek-chat"
        context: 128000
        tier: standard
        pricing: { input: 0.14, output: 0.28 }
      - model: "deepseek-reasoner"
        context: 128000
        tier: standard
        pricing: { input: 0.55, output: 2.19 }

  # ─── Mistral ─────────────────────────────────────────────────────
  mistral:
    label: "Mistral AI"
    default_model: "mistral-large-latest"
    models:
      - model: "mistral-large-latest"
        context: 128000
        tier: standard
        pricing: { input: 2.0, output: 6.0 }
      - model: "mistral-small-latest"
        context: 128000
        tier: fast
        pricing: { input: 0.20, output: 0.60 }
      - model: "codestral-latest"
        context: 256000
        tier: standard
        pricing: { input: 0.30, output: 0.90 }

  # ─── NVIDIA NIM ─────────────────────────────────────────────────
  nvidia:
    label: "NVIDIA NIM"
    default_model: "nvidia/llama-3.3-nemotron-super-49b-v1"
    models:
      - model: "nvidia/llama-3.3-nemotron-super-49b-v1"
        context: 131072
        tier: reasoning
      - model: "nvidia/nemotron-3-super-120b-a12b"
        context: 1000000
        tier: reasoning
      - model: "meta/llama-3.3-70b-instruct"
        context: 131072
        tier: standard
      - model: "meta/llama-4-maverick-17b-128e-instruct"
        context: 1000000
        tier: standard
      - model: "deepseek-ai/deepseek-v4-flash"
        context: 64000
        tier: reasoning

  # ─── Groq ────────────────────────────────────────────────────────
  groq:
    label: "Groq"
    default_model: "llama-3.3-70b-versatile"
    models:
      - model: "llama-3.3-70b-versatile"
        context: 131000
        tier: standard
        pricing: { input: 0.59, output: 0.79 }
      - model: "mixtral-8x7b-32768"
        context: 32000
        tier: standard
        pricing: { input: 0.24, output: 0.24 }

  # ─── OpenRouter ──────────────────────────────────────────────────
  openrouter:
    label: "OpenRouter"
    default_model: "openai/gpt-5.4"
    models:
      - model: "anthropic/claude-opus-4.6"
        context: 1000000
        tier: reasoning
      - model: "anthropic/claude-sonnet-4.5"
        context: 1000000
        tier: standard
      - model: "anthropic/claude-haiku-4.5"
        context: 200000
        tier: fast
      - model: "openai/gpt-5.4"
        context: 1000000
        tier: standard
      - model: "openai/gpt-5.4-pro"
        context: 1000000
        tier: reasoning
      - model: "openai/gpt-5.4-nano"
        context: 400000
        tier: fast
      - model: "openai/gpt-5.4-mini"
        context: 400000
        tier: fast
      - model: "openai/gpt-5.3-codex"
        context: 400000
        tier: standard
      - model: "google/gemini-2.5-pro"
        context: 1000000
        tier: standard
      - model: "google/gemini-2.5-flash"
        context: 1000000
        tier: fast
      - model: "x-ai/grok-4.20-beta"
        context: 2000000
        tier: standard
      - model: "moonshotai/kimi-k2.5"
        context: 262000
        tier: standard
      - model: "qwen/qwen3.5-plus-02-15"
        context: 1000000
        tier: standard
      - model: "qwen/qwen3.5-35b-a3b"
        context: 1000000
        tier: fast
      - model: "stepfun/step-3.5-flash"
        context: 1000000
        tier: fast
      - model: "z-ai/glm-5"
        context: 80000
        tier: standard
      - model: "z-ai/glm-5-turbo"
        context: 80000
        tier: fast
      - model: "minimax/minimax-m2.7"
        context: 204000
        tier: fast
      - model: "minimax/minimax-m2.5"
        context: 204000
        tier: fast
      - model: "nvidia/nemotron-3-super-120b-a12b"
        context: 262000
        tier: standard
      - model: "nvidia/nemotron-3-super-120b-a12b:free"
        context: 262000
        tier: standard
      - model: "arcee-ai/trinity-large-preview:free"
        context: 128000
        tier: standard
      - model: "xiaomi/mimo-v2-pro"
        context: 1050000
        tier: reasoning

  # ─── Ollama (local) ──────────────────────────────────────────────
  ollama:
    label: "Ollama (local)"
    default_model: "gemma4:latest"
    models:
      - model: "gemma4:latest"
        context: 128000
        tier: standard
      - model: "llama3.2"
        context: 128000
        tier: standard
      - model: "codellama"
        context: 16000
        tier: standard
      - model: "mistral"
        context: 32000
        tier: standard
      - model: "phi3"
        context: 128000
        tier: fast
      - model: "qwen2.5-coder"
        context: 128000
        tier: standard

  # ─── LM Studio (local) ──────────────────────────────────────────
  lmstudio:
    label: "LM Studio (local)"
    default_model: "default"
    models:
      - model: "default"
        context: 128000
        tier: standard