Skip to content
AI Primer

Claude Opus 4.8

Hybrid reasoning model built for serious coding and AI agents, featuring a 1M context window

A specific Claude Opus model release from Anthropic, positioned as a language model.

Pricing

Official site · Sep 23, 2026, 9:12 AM
Input / 1M
$5.00
Output / 1M
$25.00
Cached input / 1M
$0.50

Standard Claude API token pricing. Prompt caching: 5-minute cache writes $6.25/MTok; 1-hour cache writes $10/MTok; cache hits and refreshes $0.50/MTok.

Anthropic's current official pricing page lists Claude Opus 4.8 at $5 per million input tokens and $25 per million output tokens. The same page lists prompt-caching rates of $6.25/MTok for 5-minute writes, $10/MTok for 1-hour writes, and $0.50/MTok for cache hits.

View source

Model Intelligence

Arena ranking
42
Benchmarkable
Yes
Model level
release
Intelligence Index
42
Coding Index
74.3
GPQA
0.92
HLE
0.49
SciCode
0.54
IFBench
0.62
LCR
0.78
TerminalBench Hard
0.58
TAU2
0.94

Recent stories

4 linked stories
newsSECONDARY2026-06-12
Anthropic removes Fable 5 and Mythos 5 access after US directive

Anthropic said a US government directive forced it to disable Fable 5 and Mythos 5 across Claude products and APIs. The change also pushed Build Day and downstream tooling to Opus 4.8, breaking active Fable sessions and triggering fallbacks in tools like Linear Agent.

newsSECONDARY2026-06-10
Anthropic introduces visible Fable 5 fallback after Mythos complaints

Anthropic said flagged Fable 5 requests will now visibly fall back to Opus 4.8, and API refusals will return reasons instead of silently degrading output. The update matters because users were reporting sudden quality drops, opaque refusals, and quota-burn confusion around Mythos-class safeguards.

workflowPRIMARY2026-05-30
Claude Opus 4.8 supports browser-game demos in Angry Birds and GTA-style tests

Creators shared browser-game workflows that pair Magnific asset generation or single Claude prompts with playable HTML demos. The examples matter because they turn vibe-coded mini-games into short, template-driven production recipes rather than one-off experiments.

releasePRIMARY2026-05-30
Claude Opus 4.8 breaks modified-thinking flows with 400 errors

Fresh Hacker News comments and user posts report 400 errors tied to modified thinking blocks, short cache TTLs, and behavior some testers call a regression from 4.7. The issue matters because the same model is also powering one-prompt sites and games, so quota burn and client breakage are showing up alongside stronger creative output.

AI PrimerAI Primer

Your daily guide to AI tools, workflows, and creative inspiration.

© 2026 AI Primer. All rights reserved.