Skip to content

chore: sync models with dashboard API (v1.0.0) - #7

Merged
amaan-ai20 merged 1 commit into
mainfrom
model-sync/2026-09-16-1216
Sep 16, 2026
Merged

amaan-ai20 merged 1 commit into
mainfrom
model-sync/2026-09-16-1216

Conversation

@amaan-ai20

Copy link
Copy Markdown
Collaborator

Automated model-catalog sync. The dashboard API is the source of truth; every value below was taken from it. Merging publishes langchain-zerogpu v1.0.0 to PyPI — the release tags it and uses the CHANGELOG section below as the release notes.

Version

0.2.4 → 1.0.0 (major: ZeroGPUFollowUpQuestionsTool is removed, which breaks anyone importing it)

Corrected

Tool Model Field Was Now
ZeroGPUReasonTool gpt-oss-120b parameters 117B 120B
ZeroGPUReasonMultilingualTool qwen3-30b-a3b-fp8 parameters 30.5B 30B
ZeroGPUReasonLongContextTool glm-5.2 context window 1M / 1,048,576 262K
ZeroGPUReasonLongContextTool glm-5.2 cost vs. zerogpu_reason roughly twenty times roughly six to seven times
ZeroGPUReasonCodeTool deepseek-v4-flash-0731 routing hint "prefer it when the task is code or tool-use rather than sheer input size" "the largest context on the platform … or for input too large even for that tool"

The cost multiplier follows from the payload's own prices: glm-5.2 is $1.10/$3.50 per 1M input/output tokens against gpt-oss-120b's $0.15/$0.60 — 7.3× on input and 5.8× on output, so "six to seven times" covers both rather than stating each. glm-5.2 is still the most capable and priciest model in the catalog, so those superlatives stand.

The routing hint had to be rewritten rather than renumbered: at 262K, glm-5.2 no longer has the largest context window — deepseek-v4-flash-0731 does, at 1,048,576 — so telling agents to send large inputs to the long-context tool was backwards. The hint now points the other way and names no second-hand token count, which also keeps the audit from reading the number as a claim about deepseek.

Renamed

Tool Was Now
ZeroGPUReasonCodeTool deepseek-v4-flash deepseek-v4-flash-0731

Updated in the MODEL_REASON_CODE constant, the class docstring, the README tools table, and the EXPECTED_MODELS pin. The model's 284B parameters and 1M-token context window are unchanged; the endpoint (Chat Completions) is unchanged.

Removed

  • ZeroGPUFollowUpQuestionsTool (zerogpu_followup_questions) — the API no longer returns zlm-v1-followup-questions-edge, so the tool was already failing on every request. Gone with it: the MODEL_FOLLOWUP constant, the _questions helper it alone used, tests/unit_tests/test_questions.py, and its standard/integration test classes. Importers must drop the class; ZeroGPUToolkit.get_tools() now returns 16 tools instead of 17, and every "seventeen"/== 17 count in the package, README, and tests moved to sixteen.

Not changed

  • API models with no tool: zlm-v1-moderation-edge, llama-guard-4-12b, all-minilm-l6-v2, zlm-v1-signal-extract, bge-small-en-v1.5. The sync never creates a tool — wrapping these is a product decision.
  • 13B active per token on deepseek-v4-flash-0731 and 100+ languages on qwen3-30b-a3b-fp8: the API states neither count, but its descriptions still call the first a mixture-of-experts model and the second multilingual, so both were left as they stand rather than invented or dropped.

CHANGELOG

## [1.0.0] - 2026-09-16

Model catalog sync. `ZeroGPUFollowUpQuestionsTool` is gone — the ZeroGPU API no
longer serves the model behind it, so every call it made was already failing.
`ZeroGPUReasonCodeTool` follows its model to a version-pinned id, and the
long-context tool had been advertising a context window four times larger than
`glm-5.2` actually has, which was steering agents toward the most expensive
model on the platform for inputs that never needed it. Tool names and outputs
are unchanged apart from the notes below.

### Changed

- `ZeroGPUReasonCodeTool` now calls `deepseek-v4-flash-0731`. The API renamed
  `deepseek-v4-flash`, and the old id no longer resolves. The model's 284B
  parameters and 1,048,576-token context window are unchanged.
- `ZeroGPUReasonLongContextTool`: `glm-5.2`'s context window is 262K, not 1M.
  Corrected in the tool description agents route on, its docstring, and the
  README. It remains the most capable and priciest ZeroGPU model, but at roughly
  six to seven times `zerogpu_reason` per token rather than twenty. Because
  `deepseek-v4-flash-0731` now has the largest context window in the catalog,
  `ZeroGPUReasonCodeTool` no longer defers to this tool on sheer input size.
- `ZeroGPUReasonTool`: `gpt-oss-120b` has 120B parameters, not 117B.
- `ZeroGPUReasonMultilingualTool`: `qwen3-30b-a3b-fp8` has 30B parameters, not
  30.5B.

### Removed

- `ZeroGPUFollowUpQuestionsTool` (`zerogpu_followup_questions`).
  `zlm-v1-followup-questions-edge` is no longer served by the ZeroGPU API, so
  the tool failed on every call. Remove it from your imports; the toolkit now
  returns sixteen tools instead of seventeen.

Verification

  • audit-models.py --strict — clean apart from the five NOTE lines above
  • make lint, make mypy, make test pass (131 unit tests)
  • ZeroGPUToolkit(...).get_tools() returns 16
  • version above main's, uv lock --check clean, and the top CHANGELOG section matches it

🤖 Generated with Claude Code

https://claude.ai/code/session_01JhoWo1yb6MyxPX4QTD1VgZ


Generated by Claude Code

- ZeroGPUReasonCodeTool: deepseek-v4-flash -> deepseek-v4-flash-0731 (constant, docstring, README, test pin)
- ZeroGPUReasonLongContextTool: glm-5.2 context 1M -> 262K; cost multiplier twenty -> six to seven times zerogpu_reason (description, docstring, README)
- ZeroGPUReasonCodeTool: routing hint rewritten -- deepseek-v4-flash-0731 now has the largest context in the catalog
- ZeroGPUReasonTool: gpt-oss-120b parameters 117B -> 120B (docstring)
- ZeroGPUReasonMultilingualTool: qwen3-30b-a3b-fp8 parameters 30.5B -> 30B (docstring)
- remove ZeroGPUFollowUpQuestionsTool: zlm-v1-followup-questions-edge no longer served
- tool counts seventeen -> sixteen; version 0.2.4 -> 1.0.0; CHANGELOG section added

Source: https://api-dashboard.zerogpu.ai/api/models

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JhoWo1yb6MyxPX4QTD1VgZ
@coderabbitai

coderabbitai Bot commented Sep 16, 2026

Copy link
Copy Markdown

Important

  • 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: 0573ebbf-547e-4f3c-beb1-7b84f4587b78


Comment @coderabbitai help to get the list of available commands.

@amaan-ai20
amaan-ai20 merged commit d57d960 into main Sep 16, 2026
6 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.

2 participants