Better default max_tokens values

[simonw]

Right now those are 8192 for every model, but that just broke a pelican rendering on max effort for Opus 4.8.

[simonw]

Here's how to find them the max token values for each model using the /v1/models API:

curl https://api.anthropic.com/v1/models \
    -H 'anthropic-version: 2023-06-01' \
    -H "X-Api-Key: $(llm keys get anthropic)" | jq '.data | map({id, display_name, max_input_tokens, max_tokens})'

Output:

[
  {
    "id": "claude-opus-4-8",
    "display_name": "Claude Opus 4.8",
    "max_input_tokens": 1000000,
    "max_tokens": 128000
  },
  {
    "id": "claude-opus-4-7",
    "display_name": "Claude Opus 4.7",
    "max_input_tokens": 1000000,
    "max_tokens": 128000
  },
  {
    "id": "claude-sonnet-4-6",
    "display_name": "Claude Sonnet 4.6",
    "max_input_tokens": 1000000,
    "max_tokens": 128000
  },
  {
    "id": "claude-opus-4-6",
    "display_name": "Claude Opus 4.6",
    "max_input_tokens": 1000000,
    "max_tokens": 128000
  },
  {
    "id": "claude-opus-4-5-20251101",
    "display_name": "Claude Opus 4.5",
    "max_input_tokens": 200000,
    "max_tokens": 64000
  },
  {
    "id": "claude-haiku-4-5-20251001",
    "display_name": "Claude Haiku 4.5",
    "max_input_tokens": 200000,
    "max_tokens": 64000
  },
  {
    "id": "claude-sonnet-4-5-20250929",
    "display_name": "Claude Sonnet 4.5",
    "max_input_tokens": 1000000,
    "max_tokens": 64000
  },
  {
    "id": "claude-opus-4-1-20250805",
    "display_name": "Claude Opus 4.1",
    "max_input_tokens": 200000,
    "max_tokens": 32000
  },
  {
    "id": "claude-opus-4-20250514",
    "display_name": "Claude Opus 4",
    "max_input_tokens": 200000,
    "max_tokens": 32000
  },
  {
    "id": "claude-sonnet-4-20250514",
    "display_name": "Claude Sonnet 4",
    "max_input_tokens": 1000000,
    "max_tokens": 64000
  }
]