Under the list
Every line is metered 50% below what the provider lists. No tiers, no minimums, no monthly floor.
Your fast & cheap AI gateway with GLM 5.2
and 25 other open models
One endpoint in front of every open model worth running, in OpenAI or Anthropic format. Keep the client you already use, swap the base URL and the key, and pay per token — always below what the provider charges you directly.
base url
https://universalvoronez.tech/openai/v1 DeepSeek V3.1 DeepSeek V4 Flash DeepSeek-V4-Flash-0731 DeepSeek V4 Pro Gemma 4 31B Granite 4.1 8B Mellum2 12B A2.5B Llama 3.1 70B Llama 3.1 8B Llama 3.3 70B MiniMax M2.5 MiniMax-M3 Kimi K2.6 Kimi K2.7 Code Nemotron 3 Super 120B A12B Nemotron 3 Ultra 550B A55B GPT-OSS 120B GPT-OSS 20B Qwen3 14B Instruct Qwen3 30B A3B Instruct Qwen3 Coder 480B A35B Qwen3.5 35B A3B Qwen3.6 27B Qwen3.6 35B A3B GLM 5.1 GLM 5.2
Either wire
OpenAI or Anthropic format — one pool, one key. Swap the base URL and your client is done.
Whole rack
GLM, DeepSeek, Kimi, Qwen, Llama and the rest behind a single key, with reasoning-effort variants.
One command
writes the config of every client it finds$
curl -fsSL https://universalvoronez.tech/install.sh | sh Scans the machine, ticks the clients it finds and writes their configs. Your shell stays untouched
unless you pick the shell target. Undo with --uninstall.
opencode, Crush, Zed Full provider written into their own config, every model plus effort variants.
automaticClaude Code, Codex, Droid Endpoint and key land in settings.json, config.toml and config.json. Pick the model in the installer.
automaticContinue, Aider, Qwen, Goose, llm Config files written per client — nothing is exported into your shell.
automaticCursor, Windsurf, Cline, Roo Code, Kilo Code Paste the base URL and key as an OpenAI-compatible provider — same for JetBrains AI.
paste urlModel rack
pick a line — rates and limits follow it your app
- in / 1M
- $0.6950
- $1.3900
- out / 1M
- $2.2000
- $4.4000
- context
- 262k
- text only
- effort
- nonelowmediumhighxhighmax
patched to GLM 5.2