Use Strix with Abliteration

Use Strix with your ak_... API key and an abliteration/ model name to test source code, web applications, APIs, and infrastructure.

Prerequisites

Docker running locally
An ak_... key from the console

Install

Install the Strix CLI:

curl -sSL https://strix.ai/install | bash

Configure

Use Abliterated Large v2 for text-based assessments:

export LLM_API_KEY="ak_YOUR_API_KEY"
export STRIX_LLM="abliteration/abliterated-model-large-v2"
unset LLM_API_BASE OPENAI_API_BASE OPENAI_BASE_URL \
  LITELLM_BASE_URL OLLAMA_API_BASE

The abliteration/ prefix selects the built-in LiteLLM provider. No custom base URL or context fallback is needed when LiteLLM has current model metadata. Clear any base URL overrides left over from other providers.

Tested with Strix 1.7.0 using LiteLLM 1.101.0 and 1.106.4 with the current model catalog.

Run a test

Test a local project:

strix --target ./your-app

Test a deployed application:

strix --target https://staging.example.com

For an initial non-interactive run with bounded turns:

strix --non-interactive \
  --scan-mode quick \
  --max-turns 50 \
  --target https://staging.example.com

Only test systems you own or have explicit authorization to assess.

Choose a model

Model	Use it for	Context
`abliterated-model-large-v2`	Source code, APIs, and text-only assessments	`1000000`
`abliterated-model-large-flash`	Screenshots with a larger context window	`1048576`
`abliterated-model`	Screenshots and other image inputs	`262144`
`abliterated-model-large`	Deprecated; existing integrations only	`1000000`

To use Abliterated Large Flash:

export STRIX_LLM="abliteration/abliterated-model-large-flash"

Abliterated Large v2 and Abliterated Large are text-only. If an agent needs to inspect screenshots with Strix’s view_image tool, use abliterated-model-large-flash or abliterated-model. See models for the complete capability and context-window comparison.

Reasoning and usage limits

STRIX_REASONING_EFFORT is not forwarded to Abliteration in the tested versions. Each model uses its default reasoning level; see reasoning effort. With current model metadata, Strix reports estimated token costs. Add --max-budget 1 to limit a trial run’s estimated spend to $1. Estimates can differ from your bill because of web search charges and subscription discounts. Keep a turn limit and monitor usage in the console. Strix may warn that the model is not a recommended frontier model. This warning does not prevent the scan from running.

Troubleshooting

Authentication fails

Confirm that LLM_API_KEY contains an active ak_... key and that STRIX_LLM starts with abliteration/. Rerun Configure to clear LLM_API_BASE and its aliases: OPENAI_API_BASE, OPENAI_BASE_URL, LITELLM_BASE_URL, and OLLAMA_API_BASE.

Model limits or costs are missing

Allow LiteLLM to refresh its model catalog, then restart Strix. Older catalogs may lack context limits and pricing. If estimated costs show $0, use --max-turns and monitor Console usage instead of relying on --max-budget. If you must use an older catalog, set STRIX_CONTEXT_FALLBACK_TOKENS to the chosen model’s context from the table above. For Large v2:

export STRIX_CONTEXT_FALLBACK_TOKENS="1000000"

Image requests return `400`

Abliterated Large v2 and Abliterated Large are text-only. Switch to abliteration/abliterated-model-large-flash or abliteration/abliterated-model.

Requests return `429`

Strix can run several agents concurrently. Check your current limits in the console and see rate limits for retry behavior and available tiers.

Integrations

Security

Agents

AI testing and red teaming

SDKs & frameworks

API routing & provider switching

Runtimes

Use Strix with Abliteration

Prerequisites

Install

Configure

Run a test

Choose a model

Reasoning and usage limits

Troubleshooting

Authentication fails

Model limits or costs are missing

Image requests return `400`

Requests return `429`

​Prerequisites

​Install

​Configure

​Run a test

​Choose a model

​Reasoning and usage limits

​Troubleshooting

​Authentication fails

​Model limits or costs are missing

​Image requests return 400

​Requests return 429

Prerequisites

Install

Configure

Run a test

Choose a model

Reasoning and usage limits

Troubleshooting

Authentication fails

Model limits or costs are missing

Image requests return `400`

Requests return `429`