Claude Sonnet 5 benchmarks

Claude Sonnet 5 is a proprietary model from Anthropic, released 30 Jun 2026. It ranks 37th of 51 on our overall leaderboard and 14th of 51 for long context. At A$5.62 per million tokens (blended), it costs 1.6× the median model we track. It writes about 73 tokens a second, and answers start after 1.4 min on average, thinking included. It can run with requests kept in Australia on AWS Bedrock in Melbourne.

#37 OF 51 OVERALLA$5.62 / 1M TOKENS1M CONTEXTUPDATED 23 SEP 2026
CONTEXT WINDOW
1,000,000 tokens
MAX OUTPUT
128,000 tokens
INPUT
text, image
WEIGHTS
Proprietary
RELEASED
30 Jun 2026
OUTPUT SPEED
73 tok/s
ANSWER STARTS AFTER
1.4 min (thinking included)
ANTHROPIC DOCS
§ 02 — RESULTS

Benchmark results.

Independent test scores are for the adaptive reasoning, max effort setting. “Among tracked” ranks Claude Sonnet 5 against the 51 models on this site; LMArena ranks run across its full leaderboard.

ALL RESULTS8 OF 8 CORE SIGNALS
Claude Sonnet 5 benchmark results
BENCHMARKRESULTAMONG TRACKEDSOURCE DETAIL
INDEPENDENT TESTS · ARTIFICIAL ANALYSIS
AA Intelligence Index38.2#28 of 51Adaptive reasoning, max effort
AA Coding Index71.5#20 of 42Adaptive reasoning, max effort
GPQA Diamond91.1%#27 of 44Adaptive reasoning, max effort
Humanity’s Last Exam41.3%#28 of 51Adaptive reasoning, max effort
AA-LCR82.0%#16 of 51Adaptive reasoning, max effort
Terminal-Bench 2.180.5%#21 of 42Adaptive reasoning, max effort
τ²-Bench (banking)37.3%#21 of 41Adaptive reasoning, max effort
SciCode54.3%#25 of 46Adaptive reasoning, max effort
BLIND HUMAN VOTES · LMARENA
LMArena Text35,301 votes · ±51461#32 of 42#51 of 402 on LMArena · 13 Sep 2026
LMArena WebDev11,511 votes · ±71538#26 of 45#31 of 129 on LMArena · 22 Sep 2026
LMArena Vision10,289 votes · ±81266#24 of 33#33 of 152 on LMArena · 13 Sep 2026
LMArena Document7,411 votes · ±81476#12 of 26#15 of 44 on LMArena · 13 Sep 2026
LMArena Search40,230 votes · ±71196#10 of 12#12 of 34 on LMArena · 24 Aug 2026
LMArena Agent30,631 sessions#9#8 of 37#9 of 46 on LMArena · 15 Sep 2026
§ 03 — COST

What Claude Sonnet 5 costs in Australian dollars.

List prices converted at the snapshot’s RBA rate, excluding GST. Reasoning models are billed for their thinking as output tokens, so heavier settings cost more than the job estimates show.

LIST PRICE
Claude Sonnet 5 price per million tokens
PER MILLION TOKENSAUDUSD LIST
Input tokensA$2.81US$2.00
Output tokensA$14.04US$10.00
Blended (3 in : 1 out)A$5.62US$4.00
EVERYDAY JOBSESTIMATES
Estimated cost of common jobs in Australian dollars
JOBCOST (AUD)MEDIAN MODEL
Customer support replyPER 1,000 REPLIESA$9.83A$6.58
Summarise a 30-page documentPER 100 DOCUMENTSA$6.74A$4.74
Agentic coding taskPER 10 TASKSA$5.33A$3.72
SETTINGSMEASURED SEPARATELY
Claude Sonnet 5 settings compared
SETTINGINTELLIGENCEA$ / 1MSPEED
Adaptive reasoning, max effort38.2A$5.6273 tok/s
Adaptive reasoning, xhigh effort34.4A$5.6272 tok/s
Adaptive reasoning, high effort31.7A$5.6265 tok/s
Adaptive reasoning, medium effort28.1A$5.6266 tok/s
Adaptive reasoning, low effort24.3A$5.6264 tok/s
Non-reasoning, high effort23.2A$5.6261 tok/s
§ 04 — AUSTRALIA

Running Claude Sonnet 5 in Australia.

It can run with requests kept in Australia on AWS Bedrock in Melbourne. How the platforms compare →

AUSTRALIAN CLOUD REGIONS
Claude Sonnet 5 availability in Australian cloud regions
PLATFORMAVAILABILITYDETAIL
AWS Bedrock · SydneyGLOBAL

AWS's model card also lists an Australia-only au. profile, but its regional summary shows global only — treat as global until they agree. Provider docs, checked 23 Sep 2026

AWS Bedrock · MelbourneIN REGION

In-region through the bedrock-mantle endpoint (Anthropic Messages API); the standard runtime endpoint uses the au. profile. Provider docs, checked 23 Sep 2026

Azure · Australia EastNOT IN AU

Provider docs, checked 23 Sep 2026

Google Vertex AI · SydneyGLOBAL

Provider docs, checked 23 Sep 2026

§ 06 — QUESTIONS

Claude Sonnet 5, answered.

Claude Sonnet 5 lists at US$2.00 per million input tokens and US$10.00 per million output tokens — A$2.81 and A$14.04 at A$1 = US$0.7123 (RBA, 22 Sep 2026), excluding GST. A typical customer support reply works out at about A$9.83 per 1,000 replies, before any reasoning tokens.

Claude Sonnet 5 can run with requests kept in Australia on AWS Bedrock in Melbourne. Availability comes from the cloud providers’ own documentation; check your provider’s current terms before relying on it for data-residency obligations.

It ranks 22nd of 42 on our coding ranking, with an Artificial Analysis Coding Index of 71.5 and an LMArena WebDev rating of 1538 (31st of 129). The current leader is Claude Fable 5.1.

It ranks 19th of 41 on our agents ranking, completing 37.3% of τ²-Bench banking customer-service tasks and 80.5% of Terminal-Bench 2.1 tasks.

Artificial Analysis measures Claude Sonnet 5 at about 73 tok/s of output, with the answer starting after 1.4 min on average once thinking time is included (adaptive reasoning, max effort setting).

Anthropic documents a context window of 1,000,000 tokens (1M), with up to 128,000 output tokens. On AA-LCR, which tests reasoning across ~100,000-token document sets, it scores 82.0%.

PUT THE COMPARISON TO WORK

Need help choosing and using AI for your business?