NOUMENOpen app
NOUMEN

ibm · Open weights

Granite 4.2 8B

Added Aug 31, 2026Open weights

Compare models

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

Loading models…

Use in the app

IBM Granite 4.2 8B is an Apache 2.0-licensed dense model with native step-by-step reasoning and specialized training for agentic work. It can plan before acting, sequence tools, navigate codebases, work in terminals, and verify results across coding, search, mathematics, science, and complex instruction-following tasks.

  • Reasoning
  • Tools
  • Structured output
Input
Text
Output
Text
Reasoning effort
none · low · high
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.10
Output
$0.15
Cache read
$0.05

Reference prices as published by the model provider. API example

Specifications

Context window
131.1K
Max output
—
Average output (7 days)
2.4K tokens
Longer than 94% of 468 models

Side by side

Granite 4.2 8B 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…

Granite 4.2 8BibmThis modelQwen3 Next 80B A3B (Thinking)QwenRemove
ProvideribmQwen
Discovery tierOpen weightsOpen weights
AddedAug 31, 2026Jan 1, 2024
Limits
Context window131.07K256KBest
Max output—32.77K
Average output2.4K tokens—
Pricing, USD per 1M tokens
Input$0.10Best$0.15
Output$0.15Best$0.65
Cache read$0.05—
Performance
Request success (24h)100.0%—
Benchmark indexes
Intelligence Index11.111.2Best
Coding Index22.4Best17.4
Math Index—84.3
Capabilities
ReasoningYesNo
ToolsYesNo
Structured outputYesNo
Reasoning effortnone, low, high—

Public evaluations

Benchmarks

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

  • Granite 4.2 8B
  • Qwen3 Next 80B A3B (Thinking)

Intelligence Index

  • Granite 4.2 8B11.1
  • Qwen3 Next 80B A3B (Thinking)11.2

Granite 4.2 8B: Better than 23% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Granite 4.2 8B: 22.4Top
    • Qwen3 Next 80B A3B (Thinking): 17.4

  • Math Index

    Composite of the math evaluations

    • Granite 4.2 8B: —
    • Qwen3 Next 80B A3B (Thinking): 84.3

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Granite 4.2 8B: 2.1%
    • Qwen3 Next 80B A3B (Thinking): —

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Granite 4.2 8B: 0.3%
    • Qwen3 Next 80B A3B (Thinking): —

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Granite 4.2 8B: 310 Elo
    • Qwen3 Next 80B A3B (Thinking): —

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Granite 4.2 8B: 491 EloTop
    • Qwen3 Next 80B A3B (Thinking): 128 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Granite 4.2 8B: 0.6%
    • Qwen3 Next 80B A3B (Thinking): —

  • AA-LCRv1.1

    Reasoning across long documents

    • Granite 4.2 8B: 45.0%
    • Qwen3 Next 80B A3B (Thinking): 63.7%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Granite 4.2 8B: 9.7%
    • Qwen3 Next 80B A3B (Thinking): 12.6%Top

  • GPQA Diamond

    Graduate-level science questions

    • Granite 4.2 8B: 63.1%
    • Qwen3 Next 80B A3B (Thinking): 75.9%Top

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Granite 4.2 8B: —
    • Qwen3 Next 80B A3B (Thinking): 82.4%

  • IFBench

    Following precise output instructions

    • Granite 4.2 8B: —
    • Qwen3 Next 80B A3B (Thinking): 60.7%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Granite 4.2 8B: 0.0%
    • Qwen3 Next 80B A3B (Thinking): —

  • SciCode

    Scientific computing in Python

    • Granite 4.2 8B: 31.5%
    • Qwen3 Next 80B A3B (Thinking): —

  • LiveCodeBench

    Recent competitive programming problems

    • Granite 4.2 8B: —
    • Qwen3 Next 80B A3B (Thinking): 78.4%

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • Granite 4.2 8B: —
    • Qwen3 Next 80B A3B (Thinking): 84.3%

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Granite 4.2 8B: —
    • Qwen3 Next 80B A3B (Thinking): 9.8%

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Granite 4.2 8B: —
    • Qwen3 Next 80B A3B (Thinking): 41.5%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Granite 4.2 8B: 45.0%
    • Qwen3 Next 80B A3B (Thinking): 63.7%Top

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.

Request success (24h)
100.0%
Reliable
Hourly request success · last 72 hoursLowest hour 100.0%
  • 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": "ibm-granite/granite-4.2-8b",
    "reasoning_effort": "none",
    "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="ibm-granite/granite-4.2-8b",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)