Skip to content

feat(models): add DeepSeek V4.1 Flash to the picker - #58

Merged
VickyXAI merged 1 commit into
mainfrom
feat/deepseek-v4.1-flash
Oct 20, 2026
Merged

VickyXAI merged 1 commit into
mainfrom
feat/deepseek-v4.1-flash

Conversation

@VickyXAI

Copy link
Copy Markdown
Contributor

What changed

  • blockrun/deepseek/deepseek-v4.1-flash (live 2026-10-20, $0.14/$0.28, 1M ctx) added to CHAT_MODELS and the provider template, before deepseek-chat to match the order in ClawRouter#487. README, SKILL.md and the docs/03 curated count (70 → 71) updated.
  • Hermes has no hand-maintained aliases, and no free DeepSeek entries to retire.
  • Merge order: the new id sits in POST_TOP_MODELS_ADDITIONS until ClawRouter#487 merges. After that, the "landed" assertion requires removing it; I confirmed this against the #487 branch.

No pyproject or version change.

Checks

  • pytest: 121 passed, 4 skipped
  • Mirror test with CLAWROUTER_SIBLING_REQUIRED=1 against ClawRouter origin/main: pass
  • sync-brand-numbers --check: up to date

deepseek/deepseek-v4.1-flash is live on blockrun.ai (2026-10-20, $0.14/$0.28,
1M ctx). Placed before deepseek-chat in ClawRouter top-models.json order in
CHAT_MODELS and the provider template; README, SKILL.md and the docs/03
curated count (70 -> 71) follow. Exempted in POST_TOP_MODELS_ADDITIONS until
BlockRunAI/ClawRouter#487 merges.
@VickyXAI
VickyXAI merged commit cf0398b into main Oct 20, 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