Skip to content

Switch DeepNews bot to claude-fable-5-1 - #343

Merged
CodexVeritas merged 1 commit into
mainfrom
claude/deepnews-astra-6-2364a4
Sep 9, 2026
Merged

Switch DeepNews bot to claude-fable-5-1#343
CodexVeritas merged 1 commit into
mainfrom
claude/deepnews-astra-6-2364a4

Conversation

@CodexVeritas

Copy link
Copy Markdown
Collaborator

What changed

  • METAC_ASKNEWS_DEEPNEWS (research-only DeepNews bot) now uses asknews/deep-research/high-depth/claude-fable-5-1 instead of claude-fable-5.
  • Added a cheap low-depth integration test in code_tests/integration_tests/test_asknews.py that asserts AskNews accepts the claude-fable-5-1 model and reports it back in the response.

Why

When the bot was moved to claude-fable-5 on Sep 5th 2026 (#336), AskNews did not yet accept claude-fable-5-1. It does now, so the bot is being moved to the newer model.

Reviewer notes

  • Verified live on Sep 8th 2026: a low-depth call (~18s) and the full high-depth production preset with the podcasts source (~83s) both returned clean formatted output with no stray <final_answer> / <think> tags and response.model == "claude-fable-5-1".
  • gpt-6-astra also works on the AskNews server now (the rollout issue noted in Switch DeepNews bot to claude-fable-5 and add podcasts source #336 is resolved), but claude-fable-5-1 was chosen.
  • The asknews SDK 0.14.2 DeepNewsModel Literal does not list claude-fable-5-1 yet. The SDK does not validate the value at runtime and AskNewsSearcher accepts a plain str, so this works today with no SDK change.
  • code_tests/unit_tests/test_run_bots.py and pre-commit (black, isort, ruff, typos) pass.

🤖 Generated with Claude Code

AskNews DeepNews now accepts claude-fable-5-1 (it was not available when
the bot was moved to claude-fable-5 on Sep 5th 2026).

- METAC_ASKNEWS_DEEPNEWS now uses asknews/deep-research/high-depth/claude-fable-5-1
- Add a cheap low-depth integration test asserting AskNews accepts and
  reports the claude-fable-5-1 model

Verified live on Sep 8th 2026: low-depth call (~18s) and the full
high-depth production preset (~83s) both return clean formatted output
with the model reported back as claude-fable-5-1. gpt-6-astra also works
now but claude-fable-5-1 was preferred.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@CodexVeritas
CodexVeritas merged commit 2b8bc71 into main Sep 9, 2026
2 checks passed
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