Intelliverse

Models & tiers

The only model strings that exist, what each plan tier unlocks, and how routing decides.

Model aliases

You buy a plan, not a model. The model field accepts exactly five strings — the default alias intelliverse (equivalent to omitting the field), one alias per paid tier, and intelliverse/fusion for multi-model deliberation.

AliasPlanContextRequests/minPrice
intelliverseFree8,000 tokens10Free
intelliverse/airIntelliverse Air32,000 tokens60$10/mo
intelliverse/proIntelliverse Pro128,000 tokens300$30/mo
intelliverse/pro-maxIntelliverse Pro Max200,000 tokens1000$100/mo

Note

A key can request its own tier or below — an Air key may send intelliverse/air or intelliverse, but intelliverse/pro returns 403 tier_required. Fusion (intelliverse/fusion) requires Pro or Pro Max — see the Fusion guide.

How routing works

Every request is classified by task (chat, coding, extraction, summarization, reasoning, RAG, vision) and routed to the cheapest option in your tier's pool that clears the quality bar for that task — with automatic failover, tuned by your app's routing profile and your workspace cost/quality slider. Responses report the tier alias as the model in all metadata.

Routing transparency

Here's exactly what masking does and doesn't cover. Model identifiers are masked in all metadata: response model fields, streaming (SSE) chunks, headers, error messages, and your logs all report the tier alias, never a provider or model name. Masking does notalter generated content — if you ask the model what it is, it may name itself in the completion text, and we don't rewrite completions. You buy routed capability at a price point, not anonymity of the text itself.

Prefer to see how each request was routed? Send the opt-in transparency header and the response carries an x-intelliverse-routing header with the task class, the tier capability pool consulted, and the cost band — never model names. Works on streaming and non-streaming responses.

Shell
curl -i https://router-api.intelli-verse-x.ai/v1/chat/completions \
  -H "Authorization: Bearer $INTELLIVERSE_API_KEY" \
  -H "X-Intelliverse-Transparency: routing" \
  -d '{ "messages": [{"role":"user","content":"..."}] }'

# Response header (task class + tier capability pool + cost band):
# x-intelliverse-routing: {"task_class":"coding","pool":"pro/coding","cost_band":"mid"}

Capability classes

Tiers unlock capability classes, not model lists. What your plan's pool must be able to do:

  • Fast ChatHigh-throughput conversational responses for chat and support workloads.
  • Code GenerationCode completion, refactoring, and repository-scale reasoning.
  • Structured ExtractionJSON-mode data extraction, classification, and transformation.
  • Long ContextDocuments and transcripts beyond 128k tokens.
  • Frontier ReasoningDeep multi-step reasoning for research, math, and agentic planning.
  • VisionImage understanding, OCR, chart and screenshot analysis.
  • Tool CallingReliable function calling for agents and workflows.
  • Knowledge Base (RAG)pgvector knowledge base scoped to your App ID, grounded answers with citations.
  • FusionMulti-model panel deliberation with a judge synthesis for maximum accuracy.

The per-tier capability matrix lives on /tiers; pricing and credit grants on /pricing.

Did this doc help you?
Help us improve these docs