Skip to content

feat: add Zro provider (OpenAI-compatible gateway) - #1

Open
emirsoyturk wants to merge 8 commits into
devfrom
feat/add-zro-provider
Open

feat: add Zro provider (OpenAI-compatible gateway)#1
emirsoyturk wants to merge 8 commits into
devfrom
feat/add-zro-provider

Conversation

@emirsoyturk

Copy link
Copy Markdown
Collaborator

Adds the Zro provider (zro.moonmath.ai), an OpenAI-compatible service with three open-weight models:

  • GLM-5.2 none|high|max effort, interleaved, 524K ctx / 64K out, US region
  • DeepSeek V4 Flash none|high effort, interleaved, 1M ctx / 384K out, EU region
  • Kimi K3 low|high|max effort, interleaved, 1M ctx / 1M out, US region

Adds the Zro provider (zro.moonmath.ai), an OpenAI-compatible gateway
fronting LiteLLM->SGLang, serving three open-weight models:

- GLM-5.2             none|high|max effort, interleaved, 524K ctx / 64K out
- DeepSeek V4 Flash   none|high effort, interleaved, 1M ctx / 384K out
- Kimi K3             low|high|max effort, interleaved, 1M ctx / 1M out

Pricing/model set and reasoning effort tiers sourced from the product
bundled catalog and https://zro.moonmath.ai/pricing (2026-08-11).
Interleaved reasoning via choices[...].message.reasoning_content.
- deepseek-v4-flash-0731: thinking toggle + effort low|high|max (matches first-party deepseek); add output limit
- kimi-k3: add thinking toggle; drop redundant context limit (base already 1M), keep output delta
- glm-5.2: effective effort high|max per zhipuai baseline
Use Zro's live reasoning_effort surface instead of lab toggles, add GLM-5.3 and Flash as override-only entries, and drop restated identical limits.
Match first-party effort sets, drop host-specific none/off claims, and restore GLM-5.3 Flash vision from the lab.
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