Z-AI / Models
Models & providers
Every model the gateway can route to today. To unlock a provider's models, paste its key in Providers — Z-AI uses BYOK and pays your provider directly, with no markup.
Pricing
OpenAI · 8 models
GPT-5 family + still-supported GPT-4o and o1.
Get keyopenai/gpt-5.5openai/gpt-5.4openai/gpt-5.4-minidefaultopenai/gpt-5.4-nanoopenai/gpt-4oopenai/gpt-4o-miniopenai/o1-previewopenai/o1-miniAnthropic · 7 models
Claude 4 generation plus legacy 3.5 snapshots.
Get keyanthropic/claude-opus-4-8anthropic/claude-sonnet-4-6anthropic/claude-haiku-4-5anthropic/claude-haiku-4-5-20251001pinnedanthropic/claude-3-5-sonnet-20241022anthropic/claude-3-5-haiku-20241022anthropic/claude-3-opus-20240229Google Gemini · 7 models
Gemini 3.x frontier + 2.5 stable line.
Get keygemini/gemini-3.1-progemini/gemini-3.5-flashgemini/gemini-3-flashgemini/gemini-3.1-flash-litegemini/gemini-2.5-progemini/gemini-2.5-flashgemini/gemini-2.5-flash-liteGroq · 7 models
LPU inference — extreme tokens/sec, low latency.
Get keygroq/llama-3.3-70b-versatilegroq/llama-3.1-8b-instantgroq/openai/gpt-oss-120bgroq/openai/gpt-oss-20bgroq/meta-llama/llama-4-scout-17b-16e-instructgroq/qwen/qwen3-32bgroq/moonshotai/kimi-k2-instruct-0905Mistral · 8 models
Open-weights leader. Codestral / Devstral for code.
Get keymistral/mistral-medium-latestmistral/mistral-small-latestmistral/mistral-large-latestmistral/magistral-medium-latestmistral/ministral-8b-latestmistral/ministral-3b-latestmistral/codestral-latestmistral/devstral-medium-latestTogether AI · 11 models
DeepSeek, Kimi, MiniMax, Qwen — frontier OSS at scale.
Get keytogether/deepseek-ai/DeepSeek-V4-Protogether/moonshotai/Kimi-K2.6together/MiniMaxAI/MiniMax-M2.7together/zai-org/GLM-5.1together/Qwen/Qwen3.6-Plustogether/Qwen/Qwen3.5-397B-A17Btogether/Qwen/Qwen3-235B-A22B-Instruct-2507-tputtogether/openai/gpt-oss-120btogether/openai/gpt-oss-20btogether/meta-llama/Llama-3.3-70B-Instruct-Turbotogether/deepcogito/cogito-v2-1-671bFireworks AI · 11 models
Same OSS frontier, separate serverless backend.
Get keyfireworks/accounts/fireworks/models/deepseek-v4-profireworks/accounts/fireworks/models/deepseek-v4-flashfireworks/accounts/fireworks/models/kimi-k2p6fireworks/accounts/fireworks/models/minimax-m2p7fireworks/accounts/fireworks/models/glm-5p1fireworks/accounts/fireworks/models/qwen3p6-plusfireworks/accounts/fireworks/models/gpt-oss-120bfireworks/accounts/fireworks/models/gpt-oss-20bfireworks/accounts/fireworks/models/llama-v3p3-70b-instructfireworks/accounts/fireworks/models/qwen3-235b-a22b-instruct-2507fireworks/accounts/fireworks/models/qwen3-coder-480b-a35b-instructNexula AIBOM · 1 model
Zyora AI Labs' in-house security model (AIBOM-8B). India-first.
Get keynexula/aibom-8bCustom / self-hosted endpoints
Any OpenAI-compatible endpoint (vLLM, Ollama, llama.cpp, LM Studio, sglang, Modal endpoints) works as a custom provider — paste the base_url in Providers and Z-AI will route to it.