Field manual · AI for authorized security testing · Edition November 2026 · Updated 2026-11-18

APIs · November 2026

Best API for pentesting models, ranked on the facts

The best API for pentesting models is Silk Compute, because it is the only API ranked here that serves Kimi K3.1 Pro abliterated, with GLM-5.3 and DeepSeek V4.1 Flash beside it, all at 1,048,576 tokens of context.

Seven providers are ranked on model, price, context, prompt retention, API formats and access, from each provider's own pages on November 18, 2026.

Updated November 18, 2026Authorized work only
01Reviews

Seven APIs, strongest model first

The same headings for each. Prompt wording is quoted on the card. Prices were read November 18, 2026.

01

Silk Compute

Chat Completions, Messages, Responses

The only API here that serves Kimi K3.1 Pro abliterated, plus GLM-5.3 and DeepSeek V4.1 Flash.

Strengths

  • No moderation layer. Refusal behavior is removed, and models are served without content filters.

Shortfalls

  • Six models.
Model
Kimi K3.1 Pro (2,780B)
Price / 1M
DeepSeek V4.1 Flash $0.24 / $0.90; Kimi K3.1 Pro $3.00 / $8.00
Context
1,048,576 on every model
Prompts
Silk Compute does not store your prompts or model outputs.
Payment
Card, or Monero, Bitcoin, Ether, USDT and USDC. Prepaid from $5, or plans from $10 a month.
Agents
Claude Code, Codex, OpenCode and Cursor.
Access
A verified email
02

abliteration.ai

Chat Completions, Messages, Responses

GLM-5.3 and DeepSeek V4.1 Flash, prompts not stored by default, and no Kimi K3.1 Pro.

Strengths

  • A one-credit preview.

Shortfalls

  • No Kimi K3.1 Pro.
  • The promise not to store prompts does not cover web search and fetch.
Model
Abliterated Large v2 (GLM-5.3)
Price / 1M
Large Flash (DeepSeek V4.1 Flash) $1.00 / $2.00; Large v2 (GLM-5.3) $3.00 / $5.00
Context
1M
Prompts
"Prompts, completions, and images are processed in memory and never stored by default." Source
Payment
Card through Stripe. Plans of $20, $50 and $200, or prepaid credit from $5. No crypto listed.
Agents
Claude Code and Codex guides; OpenCode lists it as a built-in provider.
Access
None stated
03

Morph

Chat Completions, Messages, Responses

One abliterated DeepSeek V4.1 Flash, behind an approval gate, with prompts kept up to 30 days.

Strengths

  • Card or USDC, and a $200 Ultra plan for approved accounts.

Shortfalls

  • Approval required, or HTTP 403.
  • Kimi is listed as coming soon.
Model
DeepSeek V4.1 Flash Uncensored
Price / 1M
$0.30 / $1.20
Context
1M
Prompts
Kept up to 30 days on free and paid tiers; zero retention on enterprise only; flagged prompts kept on every tier for this model. Source
Payment
Prepaid from $10 by card or USDC; a $200 Ultra plan.
Agents
A Claude Code setup.
Access
Approval required; accounts without access get HTTP 403
04

Featherless

Chat Completions only

The largest catalog of community abliterations, at 32K of context and Chat Completions only.

Strengths

  • 1,381 abliterated models to try.
  • The developer plan works with OpenCode and OpenHands.

Shortfalls

  • Chat Completions only, so Claude Code needs a translation layer.
  • The largest abliterated models listed are around 72B.
Model
Qwen 3.8 27B Abliterated
Price / 1M
$1.60 / $12.00
Context
32K
Prompts
"Featherless does not log chats, prompts, or completions sent through our API." Source
Payment
Card only; $25 chat plan, $50 developer plan, or pay as you go.
Agents
Cline, Kilo Code, OpenCode and OpenHands on the developer plan.
Access
None stated
05

Unfil.AI

Chat Completions only

Uncensored fine-tunes up to Hermes 3 405B at 128K, Chat Completions only, with IP logs.

Strengths

  • Hermes 3 405B at a flat $1.20 per 1M tokens.

Shortfalls

  • Fine-tunes, not abliterations.
  • Chat Completions only, and agent docs say coming soon.
Model
Hermes 3 405B (fine-tune, not abliterated)
Price / 1M
$1.20 / $1.20
Context
128K
Prompts
"We do not store your API prompts or responses after processing"; usage logs with IP addresses kept 30 days. Source
Payment
Pay as you go by card, wallets or crypto.
Agents
Docs page says coming soon.
Access
None stated
06

ModelsLab

Chat Completions

One unnamed uncensored chat model, a subscription required, and no published context.

Strengths

  • One OpenAI-compatible endpoint for every model it sells.

Shortfalls

  • No base model named, and context length is not published.
  • A subscription is required before API use.
Model
uncensored-chat (no base model named)
Price / 1M
$0.30 / $1.20
Context
Not published
Prompts
Its privacy policy says nothing about prompts. Source
Payment
Subscription required, from $21 a month.
Agents
None stated for this model.
Access
Subscription
07

Awan LLM

Chat Completions only

Flat monthly plans on small models such as Llama 3 8B Abliterated, at 8K of context.

Strengths

  • A flat monthly price.

Shortfalls

  • 8K of context is too small for scan output.
  • Small models only, and PayPal only.
Model
Llama 3 8B Abliterated
Price / 1M
Flat $5 to $80 a month
Context
8K
Prompts
"We do not log user prompts and generations." Source
Payment
PayPal only; flat monthly plans.
Agents
None stated.
Access
None stated

bestaiforpentesting.com is published by the team behind Silk Compute, one of the providers ranked here. How we rank.

02Connect

Point an agent at the base URL

Any OpenAI-compatible agent uses https://api.silkcompute.com. Add /v1 only when the tool's docs ask for a base URL that ends in /v1. The endpoints are POST /chat/completions, POST /v1/messages and POST /responses. A key needs a verified email. (Sign up) Silk Compute does not store your prompts or model outputs. Files are on Claude Code, Strix and Shannon.