Repository navigation
feat(sse): add GLM-5.3 models and effort tiers - #10896
Merged
Merged
Conversation
GLM-5.3 (released 2026-08-14) ships one upstream model id with effort as a reasoning_effort request param (low|high|max, default max); thinking can no longer be disabled. Adds glm-5.3/glm-5.3-high/glm-5.3-low to GLM_SHARED_MODELS (glm/glm-cn/glmt) and glm-5.3 base to the zai catalog. - GlmExecutor: parseGlm52Effort generalized to parseGlmEffortTier; 5.3 tiers stay on the OpenAI coding transport and inject reasoning_effort + thinking.type=enabled; 5.2 tiers keep the Anthropic effort beta path - modelSpecs + GLM_PRICING entries for the three ids - default context window (1M) and pricing mirrored from GLM-5.2 (same base model) until Z.ai publishes official 5.3 specs Extracted from #10358 (unrelated .planning/codebase/ scaffolding dump dropped).
Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
A second npm run gen:provider-reference in this worktree non-deterministically dropped the cursor 'image' capability tag and devin-cli's '(Official)' suffix (unrelated to GLM-5.3, cause not pinned down). Restored to the correct content. Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>
muhamadgalihsaputra
pushed a commit
to niyatna/NiyatnaRoute
that referenced
this pull request
Sep 27, 2026
Merged — clean extraction from cheaper-inference#10358's genuinely new content (see PR body for the rationale: an unrelated .planning/codebase/ scaffolding dump was dropped). typecheck/file-size/changelog/provider-consistency gates clean, 18/18 tests passing.
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
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.
Clean extraction of #10358's value commit (by @phuongddx). The original branch carried an extra unrelated commit —
.planning/codebase/(7 files, 2189 lines) — self-generated local-tooling scaffolding docs, out of scope for this PR's title and outside the repo's file-placement conventions. Dropped from this extraction; everything else is the original commit, unmodified, plus a regeneration ofdocs/reference/PROVIDER_REFERENCE.md(auto-generated, base-drifted since 2026-08-14).What
Adds GLM-5.3 support (released 2026-08-14, https://z.ai/blog/glm-5.3) across the z.ai first-party providers.
Upstream ships one model id
glm-5.3— effort is now a request parameter (reasoning_effort:low|high|max, defaultmax) on the coding chat/completions endpoint, andthinking.type: "disabled"is no longer accepted. Per operator decision, OmniRoute keeps the 5.2 tier UX withglm-5.3-high/glm-5.3-lowaliases (base id = upstream defaultmax).Changes
open-sse/config/glmProvider.ts—glm-5.3,glm-5.3-high,glm-5.3-lowinGLM_SHARED_MODELS(→ glm / glm-cn / glmt)open-sse/config/providers/registry/zai/index.ts— baseglm-5.3only (tiers are GlmExecutor aliases;zaisends ids verbatim)open-sse/executors/glm.ts—parseGlm52Effortgeneralized toparseGlmEffortTier: 5.2 tiers keep the Anthropic-transport effort beta path (unchanged); 5.3 tiers stay on the OpenAI coding transport and injectreasoning_effort+thinking.type=enabledsrc/shared/constants/modelSpecs.ts— specs for the 3 ids +AUTHORITATIVE_CONTEXT_WINDOW_MODEL_IDSsrc/shared/constants/pricing/shared-tiers.ts—GLM_PRICINGentries for the 3 idsdocs/reference/PROVIDER_REFERENCE.md— regenerated (npm run gen:provider-reference)Spec caveats (unpublished upstream)
glm-5.3[1m]suffix for 1M but not the default; 1M mirrored from GLM-5.2 (same base model) per operator decision — correct when the official spec lands.Validation
npm run typecheck:core— cleannode scripts/check/check-file-size.mjs/check-changelog-integrity.mjs— OKnpm run check:provider-consistency— OK (263 REGISTRY entries, 343 canonical providers, 0 exceptions)tests/unit/glm-5.3-catalog-and-effort-tiers.test.ts+tests/unit/zai-catalog-glm52.test.ts— 18/18 passingCo-authored-by: phuongddx 95doanphuong@gmail.com