Claude Opus 4.8
Legacy Engineer at Anthropic | Agentic judgment and efficient tool use | 1M-token context
The basics
Checked Oct 22- Status
- Model ID
- Price
- Context
- Supported to
Updates
Strengths and weaknesses
Strengths
Weaknesses
- !
- !
Role
Flagship model at Anthropic
Sep 2026 to now · 2 months
Jul 2026 to Sep 2026 · 2 months
May 2026 to Jul 2026 · 2 months- Jul 24, 2026: Legacy model
- May 28, 2026: Flagship model
Works with
From recorded changes
Agent · 2 shared changes
App · 1 shared change
API · 1 shared change
Agent · 1 shared change
Agent · 1 shared change
Similar models
$5 in, $25 out · Compare
$4 in, $20 out
$2 in, $10 out
Cost vs peers
Tied for most expensive of 6 similar models by output price.
Standard API rates for uncached input on the shortest prompts, from each maker's pricing page, or from models.dev for models not researched here yet. Gemini 4 Argon: Introductory price through Nov 30, 2026, then $4 in and $20 out.
Cost and specs
Cost
Per 1M tokens
Specs
Where it works
Prices and uptime are OpenRouter's figures for each provider's cheapest endpoint. Uptime covers the last 24 hours.
Benchmarks
Higher is better. Green is Claude Opus 4.8. Dark marks are similar models.
Independent
Self-reported by Anthropic
Online-Mind2Web
84%
What people say
Reviews (1)
Lior GrossmanShipping software since 1993, now building with AIOpus 4.8 was my daily driver in Claude Code around June-July 2026, writing most of my code. It was replaced by Opus 5, which has its own share of problems and annoyances. As a nonfiction book writer it lost my blind test. An AI judge preferred it, but I picked Sonnet 4.6 instead because Sonnet's chapters were plainer and more concrete.
Oct 30, 2026
Launch quotes, picked by Anthropic
From its launch post, so not a neutral sample.
"Claude Opus 4.8 has noticeably better judgment. In Claude Code, it asks the right questions, catches its own mistakes, pushes back when a plan isn’t sound, and builds up confidence around complex, multi-service explorations before making big changes. It’s a great model to build with."
"On our Super-Agent benchmark, Claude Opus 4.8 is the only model to complete every case end-to-end, beating prior Opus models and GPT-5.5 at parity on cost. For agent products in translation, deep research, slide-building, and analysis, it delivers powerful reliability."
Incidents
Incidents per week, last 13 weeks. Most in one week: 1.
- Sep 3Anthropic reported elevated errors across several Claude models2 h 57 min
- Aug 24Anthropic reported elevated errors on Claude models3 h 24 min
- Jul 9Anthropic reported elevated errors for Claude Opus 4.822 min
Elevated errors, 3 times in 2 months
ResolvedWriting habits
Creative writing · EQ-Bench, among 51 models tracked here
Slop score
13.2
Overused AI words and phrases. Less slop than 54% of tracked models.
Repetition
3.7
Repeated words and ideas. Less repetition than 54% of tracked models.
Story length
5,842
Average characters per story. Longer than 18% of tracked models.