Skip to content

chore: sync capabilities: add deepseek/deepseek-v4.1-flash - #8

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

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

Conversation

@VickyXAI

Copy link
Copy Markdown
Contributor

What changed

  • model-capabilities.ts regenerated with scripts/sync-model-capabilities.mjs from the live catalog (116 ids, 89 text models).
    • Added: deepseek/deepseek-v4.1-flash (1,048,576 ctx, 65,536 out, no vision, supportsTools: false). No other entries changed.
  • README count updated. dist/ rebuilt.

Decisions

  • deepseek-chat's tool capability and profile were not copied.
    • Every other id gets supportsTools only from its own probe result.
    • capability-overrides.json holds a few hand-reasoned exceptions (a gateway incident, catalog category gaps), and none of them copies evidence from one id to another.
    • deepseek-chat's evidence is from 2026-08-29, when it served V4 Flash, so it was never a probe of V4.1.
    • A one-model probe (--only deepseek/deepseek-v4.1-flash, cents) is the way to admit it.
  • config.ts keeps deepseek/deepseek-chat in its chains. Swapping in the new id is not a no-op:
    • It would lose tool eligibility.
    • Solana does not list deepseek/deepseek-v4.1-flash yet, so Solana hosts would route to an unlisted model.
    • Revisit once Solana lists it and a probe admits it.
  • Every chain id is still live. No nvidia/deepseek-* id appears in the chains.
  • The deepseek/deepseek-v4-pro price comments in config.ts ($0.435/$0.87) are stale (live is $1.32/$3.96). They are left untouched here because they are comments only.

Source of truth

GET https://blockrun.ai/api/v1/models, fetched 2026-10-20T04:45Z.

Checks

npm run check: typecheck passes, vitest 146/146 (decision snapshot unchanged), build passes. No version bump, tag or publish.

Regenerate model-capabilities.ts with scripts/sync-model-capabilities.mjs
from GET https://blockrun.ai/api/v1/models (116 ids, 89 text models).
deepseek/deepseek-v4.1-flash is added with a 1M context window and 64K
output cap. It is not vision-capable, and supportsTools is false (no probe
evidence).

deepseek-chat's tool capability and profile are not copied. Every other id
gets supportsTools only from its own probe result (probe-tools.json).
capability-overrides.json holds a few hand-reasoned exceptions, and none of
them copies evidence from one id to another. deepseek-chat's evidence also
dates from 2026-08-29, when it served V4 Flash, so it was never a probe of
V4.1. A one-model paid probe is the right way to admit the new id.

config.ts chains keep deepseek/deepseek-chat. Swapping in the new id would
not be a no-op: it would lose tool eligibility (fails closed), and Solana
does not list v4.1-flash yet, so Solana hosts would route to a model their
gateway does not list. Revisit once Solana lists the new id and a probe
admits it.

README count updated; dist/ rebuilt.
@VickyXAI
VickyXAI merged commit 2cd5a10 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