Claude Opus 5 vs Claude Sonnet 5

Claude Opus 5 comes out ahead on 12 of 14 shared measures, and the two rank 4 and 37 of 51 overall. Claude Sonnet 5 is 2.5× cheaper per million tokens (blended) — A$9.83 vs A$24.57 per 1,000 replies of customer support replys.

CLAUDE OPUS 5 #4CLAUDE SONNET 5 #37UPDATED 23 SEP 2026
§ 01 — AT A GLANCE

Where each one wins.

Overall

Claude Opus 5 ranks 4th of 51; Claude Sonnet 5 ranks 37th.

Coding

Claude Opus 5 ranks 3rd of 42; Claude Sonnet 5 ranks 22nd.

Agents

Claude Opus 5 ranks 3rd of 41; Claude Sonnet 5 ranks 19th.

Reasoning

Claude Opus 5 ranks 4th of 51; Claude Sonnet 5 ranks 27th.

Long context

Claude Sonnet 5 ranks 14th of 51; Claude Opus 5 ranks 19th.

Value

Claude Opus 5 ranks 18th of 50; Claude Sonnet 5 ranks 33rd.

Price

Claude Sonnet 5 is 2.5× cheaper per million tokens (blended) — A$9.83 vs A$24.57 per 1,000 replies of customer support replys.

Speed

Claude Sonnet 5 writes 1.4× faster (73 tok/s vs 51 tok/s).

Australia

Both can be used from Australia through Anthropic’s own API, which may process requests outside Australia. Claude Opus 5 can run with requests kept in Australia on AWS Bedrock in Sydney and Melbourne. Claude Sonnet 5 can run with requests kept in Australia on AWS Bedrock in Melbourne.

§ 02 — EVERY MEASURE

Side by side.

“Level” means the gap is inside LMArena’s confidence interval, or the scores are identical. Independent test scores are for each model’s headline setting. Both models can be used from Australia through their makers’ APIs; the Australian rows show whether requests can also be processed in Australia.

BENCHMARKS, SPEED AND AVAILABILITY23 SEP 2026
Claude Opus 5 vs Claude Sonnet 5: benchmark results
BENCHMARKCLAUDE OPUS 5CLAUDE SONNET 5LEADS
AA Intelligence Index50.838.2Claude Opus 5
AA Coding Index78.071.5Claude Opus 5
GPQA Diamond93.2%91.1%Claude Opus 5
Humanity’s Last Exam54.9%41.3%Claude Opus 5
AA-LCR79.3%82.0%Claude Sonnet 5
Terminal-Bench 2.189.1%80.5%Claude Opus 5
τ²-Bench (banking)42.1%37.3%Claude Opus 5
SciCode56.4%54.3%Claude Opus 5
LMArena Text14931461Claude Opus 5
LMArena WebDev16911538Claude Opus 5
LMArena Vision12891266Claude Opus 5
LMArena Document14901476Level
LMArena Search1196
LMArena Agent#3#9Claude Opus 5
Blended price (AUD)A$14.04A$5.62Claude Sonnet 5
Output speed51 tok/s73 tok/sClaude Sonnet 5
Answer starts after34 s1.4 minClaude Opus 5
Context window1,000,000 tokens1,000,000 tokensLevel
USE FROM AUSTRALIA
Maker’s APIYESYES
Processed in AustraliaYES · CLOUDYES · CLOUD
AUSTRALIAN CLOUD REGIONS
AWS Bedrock · SydneyAU ROUTINGGLOBAL
AWS Bedrock · MelbourneIN REGIONIN REGION
Azure · Australia EastNOT IN AUNOT IN AU
Google Vertex AI · SydneyGLOBALGLOBAL
§ 03 — COST

The same jobs, in Australian dollars.

List prices at the snapshot’s RBA rate, excluding GST and reasoning tokens.

EVERYDAY JOBSESTIMATES
Estimated cost of common jobs in Australian dollars
JOBCLAUDE OPUS 5CLAUDE SONNET 5MEDIAN MODEL
Customer support replyPER 1,000 REPLIESA$24.57A$9.83A$6.58
Summarise a 30-page documentPER 100 DOCUMENTSA$16.85A$6.74A$4.74
Agentic coding taskPER 10 TASKSA$13.34A$5.33A$3.72
§ 05 — QUESTIONS

Claude Opus 5 or Claude Sonnet 5?

On our overall leaderboard Claude Opus 5 ranks 4th and Claude Sonnet 5 37th of 51. Which is better depends on the job: the table above compares each benchmark, price and Australian availability.

Claude Sonnet 5, at A$5.62 per million tokens (blended) against A$14.04 for Claude Opus 5. Reasoning-heavy settings add output tokens, so real bills depend on the effort level you run.

Both can be used from Australia through Anthropic’s own API, which may process requests outside Australia. Claude Opus 5 can run with requests kept in Australia on AWS Bedrock in Sydney and Melbourne. Claude Sonnet 5 can run with requests kept in Australia on AWS Bedrock in Melbourne.

PUT THE COMPARISON TO WORK

Need help choosing and using AI for your business?