NOUMENOpen app
NOUMEN

Anthropic · Frontier

Claude Opus 4.8 Thinking

Added May 28, 2026

Compare models

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

Loading models…

Use in the app

Anthropic Claude Opus 4.8 with thinking enabled.

  • Reasoning
  • Tools
  • Vision
  • Structured output
  • PDF input
Input
Text, Images
Output
Text
Reasoning effort
low · medium · high · xhigh · max
Discovery tier
Frontier

Pricing

USD per 1M tokens
Input
$5.00
Output
$25.00
Cache read
$0.50

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
128K
Average output (7 days)
1.3K tokens
Longer than 82% of 473 models

Side by side

Claude Opus 4.8 Thinking 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…

Claude Opus 4.8 ThinkingAnthropicThis modelGPT 5.6 TerraOpenAIRemove
ProviderAnthropicOpenAI
Discovery tierFrontierFrontier
AddedMay 28, 2026Jul 9, 2026
Limits
Context window1M1.05MBest
Max output128K128K
Average output1.3K tokens556 tokens
Pricing, USD per 1M tokens
Input$5.00$2.00Best
Output$25.00$12.00Best
Cache read$0.50$0.20Best
Cache write—$2.50
Performance
Throughput195.7 tok/s—
Time to first token2.8 s—
Request success (24h)100.0%100.0%
Benchmark indexes
Intelligence Index41.842.1Best
Coding Index74.376.7Best
Capabilities
ReasoningYesYes
ToolsYesYes
VisionYesYes
Structured outputYesYes
PDF inputYesYes
Reasoning effortlow, medium, high, xhigh, maxnone, low, medium, high, xhigh, max

Public evaluations

Benchmarks

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

  • Claude Opus 4.8 Thinking
  • GPT 5.6 Terra

Intelligence Index

  • Claude Opus 4.8 Thinking41.8
  • GPT 5.6 Terra42.1

Claude Opus 4.8 Thinking tested at max reasoning effort

Claude Opus 4.8 Thinking: Better than 86% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Claude Opus 4.8 Thinking: 74.3
    • GPT 5.6 Terra: 76.7Top

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Claude Opus 4.8 Thinking: 45.6%
    • GPT 5.6 Terra: 59.6%Top

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Claude Opus 4.8 Thinking: 17.4%
    • GPT 5.6 Terra: 24.5%Top

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Claude Opus 4.8 Thinking: 1,319 Elo
    • GPT 5.6 Terra: 1,332 EloTop

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Claude Opus 4.8 Thinking: 1,454 EloTop
    • GPT 5.6 Terra: 1,452 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Claude Opus 4.8 Thinking: 22.8%
    • GPT 5.6 Terra: 24.0%Top

  • AA-LCRv1.1

    Reasoning across long documents

    • Claude Opus 4.8 Thinking: 77.7%
    • GPT 5.6 Terra: 83.0%Top

  • MLCR-AA

    Long-context reasoning over medical records

    • Claude Opus 4.8 Thinking: 19.4%
    • GPT 5.6 Terra: 27.2%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Claude Opus 4.8 Thinking: 48.7%Top
    • GPT 5.6 Terra: 42.9%

  • GPQA Diamond

    Graduate-level science questions

    • Claude Opus 4.8 Thinking: 92.0%
    • GPT 5.6 Terra: 92.5%Top

  • IFBench

    Following precise output instructions

    • Claude Opus 4.8 Thinking: 62.2%
    • GPT 5.6 Terra: 71.2%Top

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Claude Opus 4.8 Thinking: 21.7%
    • GPT 5.6 Terra: 35.4%Top

  • SciCode

    Scientific computing in Python

    • Claude Opus 4.8 Thinking: 54.4%
    • GPT 5.6 Terra: 55.0%Top

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Claude Opus 4.8 Thinking: 58.3%Top
    • GPT 5.6 Terra: 57.6%

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Claude Opus 4.8 Thinking: 94.4%Top
    • GPT 5.6 Terra: 86.3%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Claude Opus 4.8 Thinking: 77.7%
    • GPT 5.6 Terra: 83.0%Top

Scores from Artificial Analysis, updated Oct 26, 2026. Claude Opus 4.8 Thinking was tested at max reasoning effort; compare scores measured at the same effort.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
195.7 tok/s
Faster than 84% of 256 models
Time to first token
2.8 s
Quicker than 56% of 256 models
Prompt cache hits
85.6%
About $1.15 per 1M input tokens with caching
Request success (24h)
100.0%
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 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": "anthropic/claude-opus-4.8:thinking",
    "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="anthropic/claude-opus-4.8:thinking",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Anthropic

Other Anthropic models, starting with the same family.