NOUMENOpen app
NOUMEN

anthracite · Open weights

Magnum v4 72B

Added Jan 1, 2024Open weights

Compare models

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

Loading models…

Use in the app

Upgraded model of Magnum V2 72B. From the creators of Goliath. Aimed at top-tier prose quality, trained on 55 million tokens of curated roleplay data.

  • PDF input
Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$2.01
Output
$2.99

Reference prices as published by the model provider. API example

Specifications

Context window
16.4K
Max output
8.2K
Average output (7 days)
111 tokens
Longer than 10% of 468 models

Side by side

Magnum v4 72B 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…

Magnum v4 72BanthraciteThis modelGPT-OSS 120B TEEOpenAIRemove
ProvideranthraciteOpenAI
Discovery tierOpen weightsOpen weights
AddedJan 1, 2024Jan 1, 2024
Limits
Context window16.38K131.07KBest
Max output8.19K16.38KBest
Average output111 tokens174 tokens
Pricing, USD per 1M tokens
Input$2.01$2.00Best
Output$2.99$2.00Best
Cache read—$2.00
Performance
Request success (24h)99.8%—
Benchmark indexes
Intelligence Index—11.6
Coding Index—30.4
Math Index—93.4
Capabilities
PDF inputYesNo
Reasoning effort——

Public evaluations

Benchmarks

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

  • Magnum v4 72B
  • GPT-OSS 120B TEE

Intelligence Index

  • Magnum v4 72B—
  • GPT-OSS 120B TEE11.6
3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 30.4

  • Math Index

    Composite of the math evaluations

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 93.4

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 0.2%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 0.0%

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 608 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 4.0%

  • AA-LCRv1.1

    Reasoning across long documents

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 52.0%

  • MLCR-AA

    Long-context reasoning over medical records

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 1.7%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 19.6%

  • GPQA Diamond

    Graduate-level science questions

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 78.2%

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 80.8%

  • IFBench

    Following precise output instructions

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 69.0%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 0.0%

  • SciCode

    Scientific computing in Python

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 34.0%

  • LiveCodeBench

    Recent competitive programming problems

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 87.8%

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 93.4%

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 23.5%

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 65.8%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Magnum v4 72B: —
    • GPT-OSS 120B TEE: 52.0%

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.

Prompt cache hits
0.0%
Request success (24h)
99.8%
Reliable

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.

Hourly request success · last 72 hoursLowest hour 95.8%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

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": "anthracite-org/magnum-v4-72b",
    "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="anthracite-org/magnum-v4-72b",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

anthracite

Other anthracite models, starting with the same family.