Models
Discover and pick from the catalog. Models are provider-prefixed (openai/, anthropic/, google/, deepseek/, grok/, qwen/, kimi/, minimax/, z-ai/, and more), and every model is callable through the same Spro endpoint with the same request format.
What's available
Browse the full catalog — with per-model pricing, context windows, capabilities, and serving providers — on the Models page in your dashboard.
Every provider in the catalog carries a badge telling you how it's accessed:
- Managed — Spro manages the API key for this provider, and you're billed by Spro according to your usage. Nothing to configure.
- BYOK — you can add your own API key for this provider (see BYOK). Your provider pricing, quotas, and billing relationship stay yours.
The catalog includes models from these providers:
| Provider prefix | Provider | Example model |
|---|---|---|
openai/ | OpenAI | openai/gpt-oss-120b |
anthropic/ | Anthropic | anthropic/claude-sonnet-4 |
google/ | Google AI Studio | google/gemini-2.5-flash |
deepseek/ | DeepSeek | deepseek/deepseek-reasoner |
x-ai/ | xAI (Grok) | x-ai/grok-4 |
mistralai/ | Mistral AI | mistralai/mistral-large |
moonshotai/ | Moonshot AI (Kimi) | moonshotai/kimi-k2 |
minimax/ | MiniMax | minimax/minimax-m2 |
z-ai/ | Z.AI (GLM) | z-ai/glm-5.3-flash |
perplexity/ | Perplexity | perplexity/sonar-pro |
xiaomi/ | Xiaomi | xiaomi/mimo |
qwen/ | Qwen (Alibaba) | qwen/qwen3.6-27b |
meta-llama/ | Meta Llama | meta-llama/llama-4-maverick |
thinkingmachines/ | Thinking Machines | thinkingmachines/tinker |
See the Providers page for the full list with badges — it also includes providers you can connect BYOK keys for beyond the model catalog, like Azure, Cerebras, Groq, SambaNova, Together, and custom endpoints.
New models are added continuously — the dashboard catalog is always the source of truth.
Naming conventions
Every model in Spro can be referenced in three ways:
| Form | Example | What it does |
|---|---|---|
| Default | glm-5.3-flash | Spro routes to the best available host automatically — recommended. |
| Author-prefixed | z-ai/glm-5.3-flash | The model ID as the author (Z.AI) publishes it. |
| Host-pinned | together/glm-5.3-flash | Serves the model through a specific hosting provider — Together AI here. |
For example:
glm-5.3-flash— automatic routing, failover, and the most cost-effective host for each requestz-ai/glm-5.3-flash— Z.AI's GLM model, author-prefixedtogether/glm-5.3-flash— the same GLM model served through Together AI
Note the host-pinned form: the prefix is the hosting provider (together), followed by the bare model name (no author prefix) — not together/z-ai/glm-5.3-flash.
When to pin
The default form is right for almost everyone: you get automatic routing, failover, and the most cost-effective option for each request. Pin a host only when you need a specific upstream's pricing, quotas, or deployment — e.g. together/glm-5.3-flash for Together AI's rate limits, or groq/gpt-oss-20b for Groq's latency.
