fireworks-ai: add missing low reasoning effort tier to glm-5p3 and glm-5p3-flash - #6237
Merged
rekram1-node merged 1 commit intoSep 4, 2026
Merged
Conversation
Both models accept reasoning_effort=low as a real, distinct tier on the Fireworks API — verified empirically (2026-09-04): the same prompt yields ~10/25/198 reasoning tokens for low/high/max on glm-5p3 and ~9/17/559 on glm-5p3-flash. The previous comment claimed the GLM 5.2 two-tier collapse (low/medium -> high) applied; it does not for the 5.3 family. This matches the upstream z.ai entries for glm-5.3 / glm-5.3-flash, which already list low/high/max.
Contributor
|
No actionable findings. |
github-actions
Bot
added
the
reviewer: ready
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
The Fireworks entries for GLM 5.3 and GLM 5.3 Flash list only
high/maxreasoning effort, with a comment claiming the GLM 5.2 two-tier collapse applies (low/medium→high). Downstream consumers (e.g. opencode) therefore only expose high/max for these models.That claim is incorrect for the 5.3 family.
lowis a real, distinct tier on the Fireworks API for both models. (For GLM 5.2 the collapse is real and that entry is correct as-is.)Evidence
Identical prompt against
https://api.fireworks.ai/inference/v1/chat/completions, varying onlyreasoning_effort, readingusage.completion_tokens_details.reasoning_tokens:accounts/fireworks/models/glm-5p3
accounts/fireworks/models/glm-5p3-flash
No 400s; each tier produces a clearly distinct reasoning budget. This also matches the upstream
zaiprovider entries forglm-5.3/glm-5.3-flash, which already listlow/high/max, and z.ai's docs (GLM-5.3 supports low/high/max, no medium rung).Change
glm-5p3.toml,glm-5p3-flash.toml:values = ["low", "high", "max"], corrected header comments, bumpedlast_updated.TOML parses clean (validated with Python tomllib; no bun toolchain on this machine for the full validate script — happy to run it if CI doesn't cover it).