[amaan-ai20]

Automated model-catalog sync. The dashboard API is the source of truth; every value below was taken from it. Merging publishes v3.11.0 (a minor bump — every sync is) to npm.

Supersedes #12.

Corrected

Model	Field	Was	Now
qwen3-30b-a3b-fp8	input / output per 1M	$0.05 / $0.30	$0.10 / $0.45
llama-3.1-8b-instruct-fast	input / output per 1M	$0.02 / $0.05	$0.15 / $0.28
zlm-v2-iab-classify-edge-enriched	input per 1M	$0.025	$0.05
zlm-v1-signal-extract	input / output per 1M	$0.02 / $0.05	$0.04 / $0.10
zlm-v1-moderation-edge	input / output per 1M	$0.02 / $0.05	$0.10 / $0.15
ZGPU_FALLBACK	priciest published rate	$1.10 / $3.50 (glm-5.2)	$0.40 / $1.60 (gpt-4.1-mini)

gpt-4.1-mini holds both the highest input and the highest output rate, so the falls back to a default ZeroGPU rate equality test is kept and now compares against it.

Added

Text Generation — priced, in chat --model, and in both Models tables. All seven carry a sample_responses_body, so they are routed to Responses; that route can be widened later if a Chat Completions-only confirmation appears.

Model	Context	input / output per 1M
glm-5.3-flash	1,048,576	$0.10 / $0.35
gpt-4.1-mini	1,047,576	$0.40 / $1.60
gpt-6-luna	272,000	$0.10 / $0.50
gpt-5.6-luna	272,000	$0.20 / $1.20
gpt-5.4-nano	400,000	$0.20 / $1.25
qwen3.6-35b-a3b	16,384	$0.20 / $1.50
zlm-v1-tool-decisions	24,576	$0.05 / $0.15

Other tasks — priced only; each is already callable through an endpoint command, so no new command was added.

Model	Task	input / output per 1M	Callable as
llama-3.2-3b-instruct-fast	Summarization	$0.05 / $0.30	zerogpu chat_completions -m llama-3.2-3b-instruct-fast
gliner2.5-base-v1	Text Classification	$0.02 / $0.05	zerogpu chat_completions -m gliner2.5-base-v1
gliner2.5-multi-v1	Text Classification	$0.02 / $0.05	zerogpu chat_completions -m gliner2.5-multi-v1
gliner2.5-decide-v1	Text Classification	$0.004 / $0	zerogpu chat_completions -m gliner2.5-decide-v1
whisper-tiny	Speech-to-Text	$0 / $0	zerogpu chat_completions -m whisper-tiny
chatterbox-nano	Text-to-Speech	$0 / $0	zerogpu chat_completions -m chatterbox-nano

Removed

The API no longer returns these four, so the CLI stops pricing, listing, and calling them.

• glm-5.2 — pricing, CHAT_MODELS, both Models tables, the -m glm-5.2 example, the Chat Completions routing sentences (README, DOCUMENTATION §4.3 and §5, ADDING_COMMANDS), and the fallback comment and test that named it. qwen3-30b-a3b-fp8 is now the only Chat Completions-only chat model.
• deepseek-v4-flash-0731 — pricing, CHAT_MODELS, both Models tables, its -m example, and the same routing prose.
• gliner2-base-v1 — pricing and docs; commands classify_structured, extract_entities, and extract_json deleted (files, src/cli.ts registrations, README ToC entries and sections, DOCUMENTATION rows and sections).
• gliner-multi-pii-v1 — pricing and docs; commands extract_pii and redact_pii deleted the same way.

Cascades: the README Extraction section had no commands left, so it is gone along with its ToC group; DOCUMENTATION §1 lost the three bullets naming the two removed extraction models; §4.8–4.12 were deleted and the sections after them renumbered to 4.8–4.13, with every #4NN-… anchor updated; §5's endpoint-command range is now 4.10–4.13.

Tests that used a removed id as sample data moved to a surviving model at the same rates: gliner2-base-v1 → gliner2.5-base-v1 (same task, $0.02/$0.05) in savings.test.ts and endpointCommands.test.ts, and the Chat Completions endpoint test's glm-5.2 → qwen3-30b-a3b-fp8, the model the CLI actually routes there.

Two judgement calls where the rules left slack, both chosen to match the surrounding file:

• The responses / chat_completions doc examples that posted gliner-multi-pii-v1 and gliner2-base-v1 kept their shape with the catalog's current extraction models (gliner2.5-multi-v1, gliner2.5-base-v1) rather than being deleted — the endpoint commands hold no model list, and the examples illustrate the command, not the model.
• whisper-tiny and chatterbox-nano are priced at $0/$0 with a one-line comment above them, mirroring the existing note that explains out: 0 for the embedding models.

Verification

• audit-models.py --strict — 0 findings (only NOTE lines for the non-chat models)
• npm run lint, npm run build, npm test (56 tests) pass
• zerogpu chat --help lists exactly the API's 13 Text Generation ids
• zerogpu --help shows no command whose model was removed

---

Generated by Claude Code

[coderabbitai]

Important

• 🔍 Trigger review

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

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: a4806c75-9e69-4d4d-b5d9-846b123273ff

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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