NOUMENOpen app
NOUMEN

Meta · Open weights

Llama 3.3 70B Instruct abliterated

Added Aug 8, 2025Open weights

Compare models

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

Loading models…

Use in the app

An abliterated (removed restrictions and censorship) version of Llama 3.3 70b.

Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.70
Output
$0.70

Reference prices as published by the model provider. API example

Specifications

Context window
32.8K
Max output
—
Average output (7 days)
129 tokens
Longer than 13% of 465 models

Side by side

Llama 3.3 70B Instruct abliterated 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…

Llama 3.3 70B Instruct abliteratedMetaThis modelMuse Spark 1.2MetaRemove
ProviderMetaMeta
Discovery tierOpen weightsFrontier
AddedAug 8, 2025Aug 5, 2026
Limits
Context window32.77K1MBest
Max output—65.54K
Average output129 tokens172 tokens
Pricing, USD per 1M tokens
Input$0.70Best$1.25
Output$0.70Best$4.25
Cache read—$0.15
Performance
Throughput29.4 tok/s—
Time to first token2.0 s—
Request success (24h)20.4%—
Benchmark indexes
Intelligence Index—39.6
Coding Index—72.2
Capabilities
ReasoningNoYes
ToolsNoYes
VisionNoYes
Structured outputNoYes
PDF inputNoYes
Audio inputNoYes
Video inputNoYes
Parallel toolsNoYes
Reasoning effort—minimal, low, medium, high, xhigh

Public evaluations

Benchmarks

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

  • Llama 3.3 70B Instruct abliterated
  • Muse Spark 1.2

Intelligence Index

  • Llama 3.3 70B Instruct abliterated—
  • Muse Spark 1.239.6
3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 72.2

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 40.6%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 15.7%

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 1,327 Elo

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 1,475 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 17.4%

  • AA-LCRv1.1

    Reasoning across long documents

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 79.0%

  • MLCR-AA

    Long-context reasoning over medical records

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 7.2%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 45.5%

  • GPQA Diamond

    Graduate-level science questions

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 90.4%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 7.1%

  • SciCode

    Scientific computing in Python

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 57.4%

Legacy evaluations

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Llama 3.3 70B Instruct abliterated: —
    • Muse Spark 1.2: 79.0%

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
29.4 tok/s
Faster than 19% of 247 models
Time to first token
2.0 s
Quicker than 76% of 247 models
Prompt cache hits
0.0%
Request success (24h)
20.4%

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": "huihui-ai/Llama-3.3-70B-Instruct-abliterated",
    "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="huihui-ai/Llama-3.3-70B-Instruct-abliterated",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Meta

Other Meta models, starting with the same family.