docs: sync model catalog with dashboard API

[amaan-ai20]

Automated model-catalog sync. The dashboard API (https://api-dashboard.zerogpu.ai/api/models) is the source of truth; every value below was taken from it. The API returned 14 models.

Corrected

Model	Field	Was	Now
zlm-v1-moderation-edge	max tokens	4,000	800
zlm-v2-iab-classify-edge-enriched	max tokens	400	800
glm-5.2	context window	1,048,576	262,144

Also fixed in integrations/hermes.mdx's model table (not scanned by the audit tool): deepseek-v4-flash input/output pricing was stale ($0.07/$0.14 → $0.16/$0.38), and glm-5.2's context column matched the old 1,048,576 value.

Added

(none — no models in the API are missing from the docs)

Removed

• zlm-v1-multi-iab-classify — page and playground deleted, dropped from the OpenAPI enums, docs/model-catalog.mdx, and docs/text-classification.mdx; the multilingual routing option removed from the zerogpu_classify_iab MCP tool in integrations/mcp.mdx; redirected /api-reference/models/zlm-v1-multi-iab-classify to /docs/text-classification.
• all-minilm-l6-v2 and bge-small-en-v1.5 — the API no longer returns any embedding models, so docs/embeddings.mdx and api-reference/embeddings.mdx are deleted (cascade: the task page and endpoint reference page they described are now empty), along with both model pages, both playgrounds, the /embeddings path, and the now-orphaned CreateEmbeddingRequest/Embedding/EmbeddingUsage/EmbeddingResponse schemas in api-reference/openapi/zerogpu.openapi.json. The zerogpu_embed MCP tool is removed from integrations/mcp.mdx and integrations/hermes.mdx (tool count 14 → 13). Redirected the old URLs to /docs/model-catalog.
• skills/investor-outreach/SKILL.md and its cookbook (cookbook/hermes-investor-outreach.mdx) used zerogpu_embed to rank investor candidates by cosine similarity. Since no embedding model remains, the ranking step now uses zerogpu_classify_zero_shot (deberta-v3-small) to score each candidate's thesis against the company one-liner instead — noting this substitution here since it's a judgment call, not a value taken directly from the API.

Verification

• audit-models.py --strict — 0 findings across all 14 models
• docs.json and every edited OpenAPI file parse as valid JSON
• Every /api-reference/models/<slug> link in docs.json and docs/*.mdx resolves to an existing file
• N models: counts on every Model library by task card match their listed ids

🤖 Generated with Claude Code

https://claude.ai/code/session_01VByDUXLbdehN4HnzaGgFDC

---

Generated by Claude Code

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated
zerogpu	🟢 Ready	View Preview	Sep 9, 2026, 1:23 PM

💡 Tip: Enable Automations to automatically generate PRs for you.

[coderabbitai]

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: 02638198-64b1-491c-a993-c5af67634bad

---

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