Claude Opus 4.8

May 2026
Claude Opus 4.8
Compare in
on
$4.08per task?
$5 / $25per 1M in / outcache $0.50
60 tok/sfirst token in 43 s?
1Mcontext

Rankingat max effort?

Full ranking →
1Opus 5.51,866.32Muse Wonder Full1,857.13Sonnet 5.51,837.84Muse Wonder1,770.35Fable 5.11,763.6
···
35DeepSeek V4 Pro 08131,458.136Opus 4.81,453.737Gemini 3.8 Flash1,445.4

Strong at

S
Tool calling#22 of 242Reasoning#14 of 336Science#21 of 312Knowledge#15 of 320
A
Coding#30 of 134Agents & work tasks#36 of 177Long context#75 of 314
B
Instructions#80 of 245

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperClaude Opus 5.5xhigh+26%-15%$3.46No forced tool use+5BetterMiMo v2.6 FlashthinkingOpen+11%-98%$0.062Pass back reasoning_content+3OpenDeepSeek V4.1 Flash 1017max+20%-94%$0.26Pass back reasoning_content · Chat Completions+4CheaperClaude Sonnet 5.5xhigh+19%-51%$2.01No forced tool use+5CheaperGrok 4.7xhigh+18%-21%$3.24Context: 1M → 500K+5

History

9 versions
05001,0001,5002,000Jul 26Release dateGDPval-AA 2.14.74.855.5
05001,0001,5002,000Jul 26Release dateGDPval-AA 2.14.74.855.5

Benchmarks

in Agents & Tool Use
Terminal-Bench 2.184.6#12
Terminal-Bench 4.021.7#30
Terminal-Bench Hard (version unstated)58.3#5
GDPval-AA (work tasks) 2.11,453.7#36
AA Analyst Agent45#15
AA-LCR (long context) 1.177.7#75
τ²-bench · AA94.4#22
τ³-Banking · AA34.2#30
AutomationBench-AA45.6#43
EnterpriseOps-Gym-AA44#14
AA-Briefcase 1.1 · Elo1,319.5#38
DeepSWE 1.159#20
EBR Bench28.6#14
Furniture Assembly42.5#15
Toolathlon-Verified76.2#3
τ³-bench · banking · 1.0.139.7#7
MCP Atlaslab82.2
JobBenchlab48.4
OSWorld-Verifiedlab83.4
Humanity’s Last Exam with toolslab57.9
Finance Agent v2lab53.9
SWE-bench Pro (edition unstated)lab69.2
DeepSearchQAlab84.3
DeepSWE (edition unstated)lab58

FAQ

How much does Claude Opus 4.8 cost?›

$5 per 1M input tokens and $25 per 1M output tokens; cached input $0.50 per 1M. A typical benchmark task costs about $4.08 at max effort.

Is Claude Opus 4.8 free to use?›

There is no free API; the cheapest paid price is $5 per 1M input tokens.

How good is Claude Opus 4.8, and what is it best at?›

#14 of 336 in Chat & Knowledge, #30 of 134 in Coding, #43 of 135 in Agents & Tool Use, #21 of 197 in Writing & Language, #28 of 37 in Vision & Documents. Strongest at: Coding, Agents & work tasks, Tool calling, Reasoning, Science, Long context, Knowledge.

Is Claude Opus 4.8 better than DeepSeek V4 Pro 0813?›

On GDPval-AA (work tasks) 2.1, Claude Opus 4.8 scores 1,453.7 vs 1,458.1 — lower. It is 505% more expensive per task.

How can I use Claude Opus 4.8?›

Through the API of Amazon Bedrock, Anthropic, Azure, Claude Platform on AWS, DeepInfra and 1 more.

What is the context window of Claude Opus 4.8?›

1M tokens, with up to 128K output tokens.

When was Claude Opus 4.8 released? Is there a newer version?›

Released 2026-05-28. It has been superseded by Claude Opus 5.

Is Claude Opus 4.8 open source?›

No, Claude Opus 4.8 is a closed model available through an API. The closest open-weights option here is DeepSeek V4.1 Flash 1017.

Can Claude Opus 4.8 read images?›

It accepts text and images, PDF files as input.

What is Claude Opus 4.8?›

Claude Opus 4.8 is a closed language model from Anthropic, released May 2026, the successor to Claude Opus 4.1.

What are the best alternatives to Claude Opus 4.8?›

Claude Opus 5.5 (cheaper), MiMo v2.6 Flash (better), DeepSeek V4.1 Flash 1017 (open weights).