Skip to main content
All three models are uncensored, think before answering by default (see thinking & reasoning effort), stream, and call tools.
  • abliterated-model — the general-purpose default. Multimodal (accepts image inputs, and video on Chat Completions), 256K context.
  • abliterated-model-large-v2 — the default large model. Derived from the open-weight GLM-5.3 model, further abliterated and fine-tuned by Abliteration AI. Text-only, 1M context, for harder reasoning and evaluation workloads. Runs in three reasoning modes — low, high, and max.
  • abliterated-model-large — the previous large model, derived from the open-weight GLM-5.2 model, further abliterated and fine-tuned by Abliteration AI. Text-only, 1M context.
Request a model by passing its identifier in the model field:

Limits

Context is the combined input + output budget; max output is the largest completion the model will return, reachable only when the prompt is small.

Capabilities

See compatibility matrix for the full per-endpoint breakdown.
Image and video inputs apply to abliterated-model only. abliterated-model-large and abliterated-model-large-v2 are text-only — image or video content is rejected with a 400 error on every surface.

Model listing

The live list is returned by GET /v1/models:

Pricing

Usage-based, billed on total tokens (input + output). See pricing for per-token rates by model, or the pricing page for plans and credit packs.
Last modified on August 31, 2026