Skip to content

chore(zenmux): refresh Z.AI and xAI families - #7484

Merged
rekram1-node merged 2 commits into
anomalyco:devfrom
Alcatraz-Zhang:codex/zenmux-zai-xai-families
Sep 19, 2026
Merged

rekram1-node merged 2 commits into
anomalyco:devfrom
Alcatraz-Zhang:codex/zenmux-zai-xai-families

Conversation

@Alcatraz-Zhang

@Alcatraz-Zhang Alcatraz-Zhang commented Sep 19, 2026

Copy link
Copy Markdown
Contributor

Summary

  • refresh ZenMux's live Z.AI and xAI routes against the public page allowlist and protocol model lists
  • add complete shared metadata for GLM-Image and the two Grok Voice models
  • remove six stale Grok/GLM route files that are absent from the live ZenMux catalog
  • update USD/MTok pricing, limits, modalities, and host reasoning controls using the already-reviewed chore(sync): update major ZenMux model families #7458 source snapshot

This is the second of three independent replacements for #7458 after human review requested a smaller split. It is based directly on dev at 97816c1e8, has no dependency on the other replacement branches, and changes 35 files.

Sources

Source snapshot: 2026-09-19. The protocol list APIs report effective discounted prices; non-token per-count/per-second prices are intentionally omitted from the token-cost schema.

Validation

  • split integrity: 35 expected / 35 actual files, no extras; the initial split matched reviewed chore(sync): update major ZenMux model families #7458 head d7f3703ad, and all six reviewed deletions remain deleted
  • review audit: live ZenMux name/modality checks and the GLM-5-Turbo reasoning side-channel check passed
  • bun validate
  • bun run compare:migrations
  • git diff --check
  • targeted generation tests: 10 passed; the repository-wide open-weight link inventory retains its 43 pre-existing missing-link failures, none introduced here

@Alcatraz-Zhang
Alcatraz-Zhang marked this pull request as ready for review September 19, 2026 07:00
@github-actions

Copy link
Copy Markdown
Contributor

Action items

  • [high] [possible mistake] providers/zenmux/models/z-ai/glm-4.6v-flash.toml:5 - Check: Provider display name must match the hosted route / base_model identity, not a different tier. Why: File id and base_model = "zhipuai/glm-4.6v-flash" are the standard Flash model (name = "GLM-4.6V-Flash"), but this route still publishes name = "GLM 4.6V FlashX". Elsewhere FlashX is a separate higher-priced serving option with its own id (e.g. glm-4.7-flashx, glm-5.3-flashx). Consumers will treat ZenMux z-ai/glm-4.6v-flash as FlashX. Action: Verify the live ZenMux model id/list entry. If it is Flash, rename to Flash (drop the FlashX label). If it is FlashX, rename the file/id to glm-4.6v-flashx (and only then keep the FlashX display name).
  • [medium] [possible mistake] providers/zenmux/models/z-ai/glm-4.6v-flash-free.toml:26 - Check: Host modality overrides for the same underlying model should be consistent unless the free route is a real capability delta. Why: Free Flash adds input = ["text", "image", "video", "pdf"], while paid glm-4.6v-flash inherits lab modalities without pdf, and lab glm-4.6v-flash / first-party Z.AI also omit pdf. Action: Confirm ZenMux free vs paid Flash inputs. Either add the same pdf override on the paid Flash route, or remove pdf from free if the catalog does not actually accept it.
  • [medium] [possible mistake] providers/zenmux/models/z-ai/glm-5-turbo.toml:1 - Check: interleaved should be set when this host exposes a reasoning side-channel for the model. Why: Same PR authors [interleaved] field = "reasoning_content" for ZenMux glm-5, glm-5.1, glm-5.2, glm-5.3*, and glm-5v-turbo, and first-party Z.AI / OpenRouter also mark GLM-5-Turbo with reasoning_content. Turbo is left without interleaved while gaining a toggle. Action: If ZenMux returns reasoning_content for z-ai/glm-5-turbo, add the same [interleaved] block; if it truly does not, document that host delta in the leading comment.

@Alcatraz-Zhang
Alcatraz-Zhang marked this pull request as draft September 19, 2026 07:19
@Alcatraz-Zhang

Copy link
Copy Markdown
Contributor Author

Addressed the review in 88b3129 after synchronizing dev and checking fresh Firecrawl live data. Kept the GLM 4.6V FlashX display name because the live ZenMux entry is literally slug z-ai/glm-4.6v-flash with name GLM 4.6V FlashX; added a leading host-naming note. Kept the free-route PDF delta because the live free entry advertises text,image,video,file while the paid entry advertises only text,image,video; added a leading modality note. Added interleaved.reasoning_content to GLM-5-Turbo, matching first-party/OpenRouter and same-host GLM peers. Revalidated live name/modality/side-channel checks, bun validate, compare:migrations, and diff check; targeted generation tests remain 10 pass / 1 unchanged 43-entry repository-baseline failure.

@Alcatraz-Zhang
Alcatraz-Zhang marked this pull request as ready for review September 19, 2026 07:22
@github-actions

Copy link
Copy Markdown
Contributor

No actionable findings.

@github-actions github-actions Bot added the reviewer: ready Automated review found no actionable items label Sep 19, 2026
@rekram1-node
rekram1-node merged commit bc1e1a8 into anomalyco:dev Sep 19, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

reviewer: ready Automated review found no actionable items

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants