#4 of 6 on the Abliterated API Index
Unfil.AI review: abliterated API pricing and privacy
Unfil.AI scores 35.4 and ranks #4 of 6. Fine-tuned uncensored models such as Hermes 3 405B, capped at 128K tokens of context.
CostCapabilityPrivacyAbliterated API Index
| Measure | Value | Rank | |
|---|---|---|---|
| Abliterated API Index | 35.4 | 4/6 | |
| Price | $1.20 / 1M | 3/6 | |
| Prompt retention | Deleted within 24 h | 5/6 | |
| Context window | 128K tokens | 4/6 | |
| 1M-context models | 0 models | 4/6 | |
| API formats | 4 / 10 | 4/6 | |
| Payment and access | 7 / 10 | 2/6 |
Read from Unfil.AI's own pages on November 18, 2026.Full review
At a glance
- Website
- unfil.ai
- Best model
- Hermes 3 405B
- Best-value 128K+ model
- Hermes 3 405B, $1.20 / $1.20
- API formats
- OpenAI Chat Completions
- Coding agents
- An OpenAI-compatible endpoint. Its docs page says "coming soon".
- Images
- Not stated.
- Payment
- Pay as you go, by card, wallets or crypto.
- Catalog
- About 15 models
Strengths
- Accepts crypto.
- About 15 uncensored models, including Hermes 3 405B.
Where it falls short
- The longest context is 128K tokens, an eighth of the 1M on every Silk Compute model.
- Its largest model costs $1.20 in and $1.20 out per 1M tokens, against $0.24 and $0.90 for DeepSeek V4.1 Flash at Silk Compute.
- No Anthropic Messages or Responses endpoint is documented, so Claude Code and Codex need a proxy.
- Usage logs with IP addresses are kept for 30 days.
What it says about your prompts
"We do not store your API prompts or responses after processing. Temporary processing data is deleted within 24 hours."unfil.ai/privacy-policy, read November 18, 2026
Prices
| Model | Input | Cached | Output | Context |
|---|---|---|---|---|
| Hermes 3 405B | $1.20 | None | $1.20 | 128K |
Per 1M tokens, from unfil.ai/pricing on November 18, 2026.
Visit Unfil.AI