How the Abliterated API Index works
The Abliterated API Index gives each provider up to 100 points across six measures, all read from its own pricing, documentation and privacy pages.
Price
Blended price per 1M tokens of the cheapest model with 128K of context or more, with full marks at $0.30.
Prompt retention
20 if prompts are never stored, 15 if deleted within a day, 5 if zero retention is opt-in or enterprise only, and 0 otherwise.
Context window
The longest context on any abliterated or uncensored model, scored up to 1M tokens.
1M-context models
5 points for each named abliterated model with 1M tokens of context, up to three.
API formats
OpenAI Chat Completions 4, Anthropic Messages 4 and OpenAI Responses 2.
Payment and access
Crypto 4, no subscription required 3, and a free trial or free credit 3.
Why these measures
People use abliterated models for long jobs such as agent runs and log analysis. Those need a long context, cheap input tokens and a clear prompt retention policy. Claude Code needs the Anthropic Messages format, and Codex needs Responses.
Blended price
The blend weights input 3 to 1 against output, and a provider without a 128K model scores no price points.
What we do not score
Speed and availability change by the hour and cannot be read from a page, so the index leaves them out. The model ranking covers model quality.
Sources
Who publishes this
bestabliterated.com is published by the team behind Silk Compute, one of the providers ranked here. The index scores Silk Compute with the same measures, weights and sources as every other provider.