Skip to content

Models & pricing

Inferel serves 70+ models across frontier labs and leading open-source families, including:

FamilyExamples
Anthropicclaude-opus-4.8, claude-sonnet-4.6, claude-haiku-4.5, claude-fable-5
OpenAIgpt-5.5, gpt-5.4-pro, gpt-5.4-mini, gpt-5.3-codex
Googlegemini-3.5-flash, gemini-3.1-pro-preview, gemma-4-31b-it
DeepSeekDeepSeek-V4-Pro, DeepSeek-V3.2, DeepSeek-R1-0528
QwenQwen3.7-Max, Qwen3.6-Plus, Qwen3-Coder-480B-A35B-Instruct-FP8
MoonshotKimi-K2.6, kimi-k2.7-code
OthersGLM-5.2-FP8, MiniMax-M3, MiMo-V2.5-Pro, nemotron-3-ultra-550b-a55b

The authoritative, always-current list is GET /v1/models — the catalog evolves as capacity and providers are added, and a model is listed only while it has healthy serving capacity.

LLM usage is priced per 1,000,000 tokens, with separate rates for input (prompt) and output (completion) tokens — displayed in the console as, for example, $3 / $15 per 1M tokens. Where the underlying model supports prompt caching, cache reads and cache writes are priced at their own per-1M-token rates — typically 0.1× and 1.25× the input rate respectively. See Prompt caching for how cached tokens appear in usage and exactly how they are billed.

The cost of each request is computed from the exact token counts in the response’s usage object and deducted from your organization’s prepaid credit balance — see Billing & credits.

Prices come from one source, so the number you see is always the number you are charged:

  • GET /v1/models returns each model’s live per-token rates in a pricing block. Call it with your API key and the rates already include any discount your organization has.
  • The console’s Billing page shows the same table in per-million-token form, with discounted rates highlighted alongside the list price.

Per-organization discounts and credit coupons are applied automatically — no code change is needed to pick them up.

Some models are available through multiple vendor pools with different price/performance trade-offs — see Resource groups for how to select one explicitly.