Claude Opus 5

Jul 2026
Claude Opus 5
Compare in
on
$5.86per task?
$5 / $25per 1M in / outcache $0.50
55 tok/sfirst token in 55 s?
1Mcontext

Rankingat max effort?

Full ranking →
1GPT-6 Astra742GPT-6.1 Sol683Opus 5.5674GPT-5.5635GPT-5.5 Pro62
···
22GPT-5.4 Nano (2026-03-17)4524Opus 54324GPT-5.6 Luna43

Strong at

S
Coding#9 of 132Agents & work tasks#7 of 175Reasoning#6 of 334Science#11 of 312Knowledge#7 of 318Vision#5 of 124
A
Long context#52 of 273

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperGPT-6 Astramax+72%-44%$3.26Monitoring can return HTTP 403+2BetterClaude Haiku 5.5max+28%-96%$0.21Remove sampling parameters+5OpenDeepSeek V4 Pro 0813max+7%-88%$0.67No image input+5Lower effortSame modellow-37%-81%$1.10CheaperGPT-6.1 Solmax+58%-88%$0.72Remove unsupported parameters+5

History

9 versions
0204060Jan 26Apr 26Jul 26Oct 26Release dateChess Puzzles4.54.64.74.855.5
0204060Jan 26Apr 26Jul 26Oct 26Release dateChess Puzzles4.54.64.74.855.5

Benchmarks

in Chat & Knowledge
AA Intelligence Index 4.3.250.8#9
AA Omniscience Index37.1#7
AA Omniscience · Accuracy60.9#6
AA Omniscience · Hallucination Rate60.8#107
Humanity's Last Exam54.9#6
GPQA Diamond93.2#11
AA-LCR (long context) 1.179.3#52
CritPt (physics)29.1#12
GDP.pdf · AA all-pass21.6#22
MLCR-AA55.6#6
ARC-AGI-1 · public evaluation99#2
ARC-AGI-1 · semi-private97.5#8
ARC-AGI-2 · public evaluation97.1#6
ARC-AGI-2 · semi-private90.4#6
Arena Text · overall1,490#11
Arena Text · overall · full cohort1,507.6#4
Arena Text · Chinese1,573.4#4
Arena Text · English1,507.3#5
Arena Text · French1,507.8#4
Arena Text · German1,525.9#1
Arena Text · instruction following1,517.8#4
Arena Text · Japanese1,515.4#2
Arena Text · Korean1,520.7#2
Arena Text · non-English1,500.5#4
Arena Text · Polish1,519.9#2
Arena Text · Russian1,507.2#6
Arena Text · Spanish1,519.5#2
GPQA Diamond · Epoch93.9#11
OTIS Mock AIME 2024 202598.9#15
FrontierMath Tiers 1–3 (v2)85.6#12
FrontierMath Tier 4 (v2)73.2#11
SimpleQA Verified v159.9#15
Chess Puzzles43#24
Mystery Game Puzzles59#6
FrontierMath Erdos0#6
MathArena · ArXivMath · June 202682.6#9
MathArena · BrokenArXiv · June 202690.7#5
MathArena · ArXivLean · June 202632.6#2
HLE fulllab56.3
ExploitGymlab22.1
HealthBench Professional length-adjustedlab59.8

FAQ

How much does Claude Opus 5 cost?›

$5 per 1M input tokens and $25 per 1M output tokens; cached input $0.50 per 1M. A typical benchmark task costs about $5.86 at max effort.

Is Claude Opus 5 free to use?›

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

How good is Claude Opus 5, and what is it best at?›

#6 of 334 in Chat & Knowledge, #9 of 132 in Coding, #28 of 133 in Agents & Tool Use, #10 of 195 in Writing & Language, #5 of 124 in Vision & Documents. Strongest at: Coding, Agents & work tasks, Reasoning, Science, Long context, Knowledge, Vision.

Is Claude Opus 5 better than Claude Opus 4.8?›

On Chess Puzzles, Claude Opus 5 scores 43 vs 52 — lower. It is 44% more expensive per task.

How can I use Claude Opus 5?›

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

What is the context window of Claude Opus 5?›

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

When was Claude Opus 5 released? Is there a newer version?›

Released 2026-07-24. It has been superseded by Claude Opus 5.5.

Is Claude Opus 5 open source?›

No, Claude Opus 5 is a closed model available through an API. The closest open-weights option here is DeepSeek V4 Pro 0813.

Can Claude Opus 5 read images?›

It accepts text and images, PDF files as input.

What is Claude Opus 5?›

Claude Opus 5 is a closed language model from Anthropic, released Jul 2026, the successor to Claude Opus 4.8.

What are the best alternatives to Claude Opus 5?›

GPT-6 Astra (cheaper), Claude Haiku 5.5 (better), DeepSeek V4 Pro 0813 (open weights).