Skip to content

Automated implementation via the implement pipeline - #111

Open
Ken Chau (kenotron-ms) wants to merge 3 commits into
mainfrom
resolve/amplifier-module-provider-anthropic-65e7ea80
Open

Automated implementation via the implement pipeline#111
Ken Chau (kenotron-ms) wants to merge 3 commits into
mainfrom
resolve/amplifier-module-provider-anthropic-65e7ea80

Conversation

@kenotron-ms

Copy link
Copy Markdown
Contributor

Automated contribution via the implement pipeline.


Generated by the implement pipeline.

Amplifier Resolve added 3 commits September 1, 2026 23:46
## Summary

Registers Claude Fable 5.1 in the provider module.

## Model facts (from Anthropic sources)

- **API model identifier**:
  Source: https://docs.anthropic.com/en/docs/about-claude/models/overview
  (verified 2026-09-01)
- **Context window**: 1M tokens (same as Fable 5)
  Source: https://docs.anthropic.com/en/docs/about-claude/models/overview
- **Max output tokens**: 128K tokens (same as Fable 5)
  Source: https://docs.anthropic.com/en/docs/about-claude/models/overview
- **Pricing**: $10/MTok input, $50/MTok output (same as Fable 5)
  Cache reads: $0.25/MTok (75% cheaper than Fable 5's $1.00/MTok)
  Cache writes (5-min): $12.50/MTok (same as Fable 5)
  Source: https://www.anthropic.com/claude-fable-and-mythos-5-1
  ("Cache reads now cost 75% less, or $0.25 per million tokens.")
- **New/changed API parameters**: None. Same API surface as Fable 5.

## Changes

### `amplifier_module_provider_anthropic/_cost.py`
- Added `claude-fable-5-1` to `_RATES` with the new pricing:
  - `input_per_m`: $10.00 (same as Fable 5)
  - `output_per_m`: $50.00 (same as Fable 5)
  - `cache_read_per_m`: $0.25 (75% cheaper than Fable 5's $1.00)
  - `cache_write_per_m`: $12.50 (same as Fable 5)

### `amplifier_module_provider_anthropic/__init__.py`
- Updated `_get_capabilities` docstring to reference Fable 5 / 5.1
  (capability detection is family-based, so no code change needed)

### `tests/test_fable51.py` (new)
- 27 new tests covering:
  - Family detection (`claude-fable-5-1` -> `fable`)
  - Version parsing ((5, 1) from the model ID)
  - Full capability matrix (128K output, 1M context, adaptive thinking,
    all effort levels, no speed mode, no manual thinking)
  - Cost: input/output same as Fable 5; cache_read 75% cheaper
  - Not in fast-eligible models
  - Fallback ladder: steps down to opus (same as Fable 5)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant