NOUMENOpen app
NOUMEN

abliteration · Balanced

Abliterated Model

Added Aug 27, 2026

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Abliteration.ai's unrestricted multimodal reasoning model is weight-modified to reduce refusals and answer prompts the original model might reject. It supports text and image input, structured output, automatic prompt caching, and a 262K-token context window.

  • Reasoning
  • Vision
  • Structured output
Input
Text, Images
Output
Text
Reasoning effort
none · minimal · low · medium · high · xhigh · max
Discovery tier
Balanced

Pricing

USD per 1M tokens
Input
$1.00
Output
$3.00
Cache read
$0.10

Reference prices as published by the model provider. API example

Specifications

Context window
262.1K
Max output
262.1K
Average output (7 days)
1.4K tokens
Longer than 83% of 473 models

Side by side

Abliterated Model compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Abliterated ModelabliterationThis modelStep 5 PreviewstepfunRemove
Providerabliterationstepfun
Discovery tierBalancedBalanced
AddedAug 27, 2026Sep 20, 2026
Limits
Context window262.14K1MBest
Max output262.13K—
Average output1.4K tokens1.5K tokens
Pricing, USD per 1M tokens
Input$1.00$1.00
Output$3.00$2.70Best
Cache read$0.10$0.05Best
Performance
Throughput171.9 tok/s—
Time to first token1.8 s—
Request success (24h)88.7%100.0%Best
Benchmark indexes
Intelligence Index—43.7
Capabilities
ReasoningYesYes
ToolsNoYes
VisionYesYes
Structured outputYesYes
Video inputNoYes
Parallel toolsNoYes
Reasoning effortnone, minimal, low, medium, high, xhigh, maxlow, medium, high

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Abliterated Model
  • Step 5 Preview

Intelligence Index

  • Abliterated Model—
  • Step 5 Preview43.7
3.8Catalog models by score55.1

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Abliterated Model: —
    • Step 5 Preview: 51.0%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Abliterated Model: —
    • Step 5 Preview: 17.7%

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Abliterated Model: —
    • Step 5 Preview: 1,423 Elo

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Abliterated Model: —
    • Step 5 Preview: 1,587 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Abliterated Model: —
    • Step 5 Preview: 14.8%

  • AA-LCRv1.1

    Reasoning across long documents

    • Abliterated Model: —
    • Step 5 Preview: 88.3%

  • MLCR-AA

    Long-context reasoning over medical records

    • Abliterated Model: —
    • Step 5 Preview: 18.3%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Abliterated Model: —
    • Step 5 Preview: 46.5%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Abliterated Model: —
    • Step 5 Preview: 33.3%

  • SciCode

    Scientific computing in Python

    • Abliterated Model: —
    • Step 5 Preview: 58.9%

Legacy evaluations

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Abliterated Model: —
    • Step 5 Preview: 88.3%

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Throughput
171.9 tok/s
Faster than 79% of 256 models
Time to first token
1.8 s
Quicker than 77% of 256 models
Prompt cache hits
50.3%
About $0.55 per 1M input tokens with caching
Request success (24h)
88.7%

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "abliteration-ai/abliterated-model",
    "reasoning_effort": "medium",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="abliteration-ai/abliterated-model",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

abliteration

Other abliteration models, starting with the same family.