Retire Kimi K2.6 and K2.7 Code - #7
Merged
Merged
Conversation
Both leave the catalog. Kimi K3 stays as the Kimi offering.
Two other models named kimi-k2p7-code in their declared fallback chains,
so those are repaired rather than left dangling — a chain naming a model
the catalog no longer has resolves to nothing, which silently shortens the
safety net instead of announcing itself:
glm-5p2 kimi-k2p7-code, terra -> terra
kimi-k3 kimi-k2p7-code, terra -> glm-5p2, terra
kimi-k3 keeps a same-provider hop that still exists rather than going
straight off Fireworks.
The kimi-k2p7 pricing family is removed with the model it existed for.
The generic kimi row stays as the floor, so a stale pin to a retired id
still meters at roughly its real rate instead of falling to the defaults.
A stale pin is otherwise harmless: an id outside the catalog is refused by
the picker and by policy validation, and the pin resolver falls through to
the next scope.
timothyerwin
added a commit
that referenced
this pull request
Sep 3, 2026
catalog/models.json and models.json: resolved onto main's post-#7 catalog as the base, so both retired Kimi rows stay gone, then re-applied the 14 cross-vendor chains on top. Resolved structurally against the parsed catalog rather than by editing conflict text — the file round-trips byte-identically through json, so the rewrite touches only the fallback arrays. policy_test.go: a comment-wording conflict only; took main's phrasing. No chain references a retired model, and TestFallbackFirstHopLeavesTheVendor still passes: every first hop changes vendor.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Removes Kimi K2.6 and Kimi K2.7 Code from the model catalog. Kimi K3 stays as the Kimi offering.
Why the change is more than two deletions
Both models appeared in other models' declared fallback chains. A chain naming a model the catalog no longer contains resolves to nothing — it doesn't error, it just silently shortens the safety net that keeps a turn alive when a provider is unreachable. So the chains are repaired, not left dangling:
glm-5p2kimi-k2p7-code, terraterrakimi-k3kimi-k2p7-code, terraglm-5p2, terrakimi-k3keeps a same-provider hop that still exists rather than falling straight off Fireworks on its first failure.The
kimi-k2p7pricing family is removed along with the model it was written for. The generickimirow stays as the floor, so a stale pin to a retired id still meters at roughly its real rate instead of dropping to the$3/$15defaults.What happens to someone pinned to one of them
Nothing breaks. A model id outside the catalog is refused by the
/modelpicker and by policy validation, and the pin resolver falls through to the next scope — workspace, then user, then thedefault_modelseed.Verification
gofmt,go vet,go test -race ./...,staticcheck— the full CI gate, cleanmodels.jsoncopies verified byte-identical aftergo generate ./catalogand the sync scriptfallbackentry resolves to a model still in the catalogNote
The first attempt at this used index arithmetic over the JSON text and removed the wrong entry —
qwen3p8-2p4t-a95binstead ofkimi-k2p6. Caught by diffing the parsed catalog againstHEADrather than trusting the edit. Redone as a structural edit after verifying a JSON round-trip is byte-identical to the original file.The matching www and gateway changes (catalog sync, FAQ, marketing model list, test fixtures) go direct to
maininmemcode.ai.