Models
Choose the right intelligence.
Explore representative model families and the intended routing design. Live Composer currently uses the chat ability on the tenant’s region-locked AIHK or AICA gateway.
Provider-family names, strengths, latency and cost labels are representative catalogue guidance, not a live inventory or benchmark. Outside-provider selection is not currently exposed to tenants.
Catalog
Representative provider families.
This table describes intended future catalogue grouping. It is not the current AIHK/AICA tenant ability list.
| Family | Provider | Strength | Latency | Cost | |
|---|---|---|---|---|---|
| GPT-class | OpenAI | Broad reasoning, tool use | Fast | $$ | Popular |
| Claude-class | Anthropic | Long documents, structure retention | Fast | $$ | Long context |
| Gemini-class | Multimodal, cost efficiency | Very fast | $ | Value | |
| Llama-class | Open source | Open weights, private deployment | Fast | Free tier | Open |
| Mistral-class | Mistral | Compact, low latency | Very fast | $ | Compact |
| Auto | GenCMS Router | Routes each task to the best fit | Varies | Optimised | Default |
Cost and latency indicators are relative guidance, not a quote. Actual credits depend on the model version, prompt length and output length of each run.
Routing
Not locked to one AI company.
The intended router would let you pin a model, use a Skill recommendation or set a preference for cost, latency or quality. Automatic routing is not connected yet.
Pin
Planned: keep a Skill on one selected model.
Recommend
Planned: show a model suggestion and its reasoning before you choose.
Auto
Planned: route each run against a stated preference.