Head-to-head comparison

Claude Opus 4.6 vs. Claude Opus 4.5

Compare published benchmark results from matching versions and cohorts, API pricing, and technical specifications. Any documented test setup differences remain visible.

shared result series: 15
distinct benchmarks: 10
latest retrieval: 2026-09-29

First modelSecond model

Specifications

General specifications

Release date, availability, context window, architecture, and published parameter counts at a glance. Undisclosed details remain marked as unavailable.

Specifications for Claude Opus 4.6 and Claude Opus 4.5
Attribute	Claude Opus 4.6	Claude Opus 4.5
Provider	Anthropic	Anthropic
Released	Feb 2026	Nov 2025
Availability	Active	Active
Model type	Proprietary	Proprietary
Parameters	Not published	Not published
Architecture	Not published	Not published
Context window	1,000,000 tokens	200,000 tokens
Knowledge cutoff	May 2025	May 2025
Technical sources	Model, Context	Model, Context

Pricing

API pricing compared

The table compares official standard API rates per 1M tokens, including published cache pricing. Batch, fast, priority, regional surcharges, tool calls, and cloud platform pricing are excluded.

API pricing for Claude Opus 4.6 and Claude Opus 4.5
Price type	Claude Opus 4.6	Claude Opus 4.5
API input	$5	$5
API output	$25	$25
Cache read	$0.5	$0.5
Cache write (5 min.)	$6.25	$6.25
Cache write (1 hr.)	$10	$10
Price verified	09/22/2026 Source	09/22/2026 Source

Claude Opus 4.6Claude Opus 4.5

API inputUSD per 1M tokens, base tier

Claude Opus 4.6$5

Claude Opus 4.5$5

API outputUSD per 1M tokens, base tier

Claude Opus 4.6$25

Claude Opus 4.5$25

Cache readUSD per 1M tokens, base tier

Claude Opus 4.6$0.5

Claude Opus 4.5$0.5

Key to the marks:better valuebetter, but not conclusivetie

Pricing by provider

Direct developer pricing and provider endpoint tariffs routed through OpenRouter are listed separately. Prices are in USD per 1M tokens. Context-dependent direct tiers plus regional, Flex, Priority, and other OpenRouter tariffs remain individually identifiable.

Provider pricing for Claude Opus 4.6 and Claude Opus 4.5
Model	Provider	Purchase route and tariff	Input per 1M tokens	Output per 1M tokens	Source
Claude Opus 4.6	Anthropic	Direct from developerStandard API	Lowest listed price: $5	Lowest listed price: $25	Source09/22/2026
	Amazon Bedrock	Via OpenRouteramazon-bedrock	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026
	Anthropic	Via OpenRouteranthropic	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026
	Azure	Via OpenRouterazure/global	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026
	Claude Platform on AWS	Via OpenRouterclaude-on-aws	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026
	Google	Via OpenRoutergoogle-vertex/europe	$5.5	$27.5	Source10/15/2026
	Google	Via OpenRoutergoogle-vertex/global	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026
Claude Opus 4.5	Anthropic	Direct from developerStandard API	Lowest listed price: $5	Lowest listed price: $25	Source09/22/2026
	Amazon Bedrock	Via OpenRouteramazon-bedrock	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026
	Amazon Bedrock	Via OpenRouteramazon-bedrock/eu-west-1	$5.5	$27.5	Source10/15/2026
	Anthropic	Via OpenRouteranthropic	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026
	Azure	Via OpenRouterazure/global	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026
	Claude Platform on AWS	Via OpenRouterclaude-on-aws	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026
	Google	Via OpenRoutergoogle-vertex/global	Lowest listed price: $5	Lowest listed price: $25	Source10/15/2026

Performance

Shared benchmarks

Only published results with the same benchmark version, task, metric, and comparison cohort are paired. Different reasoning levels, agents, harnesses, or output limits appear directly in the table.

Capability profile from matched benchmarks

Each axis averages directly matched benchmark families. A value of 100 means the stronger value within this pair, not a universal quality score.

Claude Opus 4.6

Claude Opus 4.5

Reasoning: 100 / 85.3 (1 benchmark family)

Agentic tasks: 100 / 55.8 (1 benchmark family)

Multimodal: 100 / 99.6 (1 benchmark family)

Cybersecurity: 96 / 100 (1 benchmark family)

Health: 100 / 94.8 (2 benchmark families)

Education: 100 / 93 (1 benchmark family)

12 matched rows from 7 benchmark families were considered. The radar shows 6 of 6 comparable categories.

Claude Opus 4.6Claude Opus 4.5

MedCodeOverall. Higher is better. No documented setup difference

Claude Opus 4.648.244 %

Claude Opus 4.545.174 %

MedScribeOverall. Higher is better. No documented setup difference

Claude Opus 4.686.738 %

Claude Opus 4.583.246 %

METR Time Horizon50% time horizon. Higher is better. No documented setup difference

Claude Opus 4.612 h (95% confidence interval 5.3 h to 60.6 h)

Claude Opus 4.54.9 h (95% confidence interval 2.7 h to 10.4 h)

MultiChallengeOverall. Higher is better. Test setup not documented as identical

Claude Opus 4.656.02% ± 1.08

Claude Opus 4.558.97% ± 0.42

MultiNRCOverall. Higher is better. No documented setup difference

Claude Opus 4.648.34% ± 3.02

Claude Opus 4.541.23% ± 2.97

PRBench FinanceOverall. Higher is better. Test setup not documented as identical

Claude Opus 4.653.28% ± 0.18

Claude Opus 4.546.16% ± 0.27

PRBench LegalOverall. Higher is better. Test setup not documented as identical

Claude Opus 4.652.27% ± 0.66

Claude Opus 4.544.21% ± 0.34

SecCodeBench-V2Generation with security hints. Higher is better. No documented setup difference

Claude Opus 4.670.65 %

Claude Opus 4.572.1 %

TutorBenchOverall. Higher is better. No documented setup difference

Claude Opus 4.653.55% ± 1.01

Claude Opus 4.549.82% ± 0.98

VISTAOverall. Higher is better. No documented setup difference

Claude Opus 4.645.48% ± 0.44

Claude Opus 4.545.32% ± 0.03

METR Time Horizon80% time horizon. Higher is better. No documented setup difference

Claude Opus 4.61.2 h (95% confidence interval 27 min to 2.8 h)

Claude Opus 4.549.4 min (95% confidence interval 20.7 min to 1.7 h)

SecCodeBench-V2Generation without hints. Higher is better. No documented setup difference

Claude Opus 4.651.05 %

Claude Opus 4.558.02 %

Key to the marks:better valuebetter, but not conclusivetie

Benchmark scores for Claude Opus 4.6 and Claude Opus 4.5
Benchmark and task	Claude Opus 4.6	Claude Opus 4.5	Source
MedCodeOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterClaude Opus 4.5: Snapshot: claude-opus-4-5-20251101	Winner48.244 %	45.174 %	Vals AI Source details Data as of: 09/26/2026
MedScribeOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterClaude Opus 4.5: Snapshot: claude-opus-4-5-20251101	Winner86.738 %	83.246 %	Vals AI Source details Data as of: 09/26/2026
METR Time Horizon50% time horizon Test details No documented difference in the test setupVersion: TH 1.1Metric: 50% time horizonScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-2026-02-05, Agent: ReAct, Harness: ReAct, Attempts: 6 independent runs per task Claude Opus 4.5: Snapshot: claude-opus-4-5-2025-11-24, Agent: ReAct, Harness: ReAct, Attempts: 6 independent runs per taskReAct agent with six independent runs per task. Time is estimated human completion duration.	Winner12 h (95% confidence interval 5.3 h to 60.6 h)	4.9 h (95% confidence interval 2.7 h to 10.4 h)	METR Source details Data as of: 05/08/2026
MultiChallengeOverall Test details Test setup not documented as identicalNo direct winnerVersion: four multi-turn challenge typesMetric: rubric pass rateScoring: Higher is betterClaude Opus 4.6: Reasoning: none, Attempts: pass@1 Claude Opus 4.5: Reasoning: thinking, Attempts: pass@1Multi-turn conversations with binary task-specific rubrics.	56.02% ± 1.08 95% confidence interval: ±1.08	Higher measured value58.97% ± 0.42 95% confidence interval: ±0.42	Scale AI Source details Data as of: 08/26/2026
MultiNRCOverall Test details No documented difference in the test setupVersion: French, Spanish, and Chinese releaseMetric: accuracyScoring: Higher is betterSettings: Reasoning: none, Attempts: pass@1Native, non-translated reasoning tasks with automatic evaluation.	Winner48.34% ± 3.02 95% confidence interval: ±3.02	41.23% ± 2.97 95% confidence interval: ±2.97	Scale AI Source details Data as of: 08/26/2026
PRBench FinanceOverall Test details Test setup not documented as identicalNo direct winnerVersion: full finance setMetric: weighted rubric scoreScoring: Higher is betterClaude Opus 4.6: Reasoning: none, Attempts: pass@1 Claude Opus 4.5: Reasoning: thinking, Attempts: pass@1600 finance tasks with task-specific rubrics and o4-mini as judge.	Higher measured value53.28% ± 0.18 95% confidence interval: ±0.18	46.16% ± 0.27 95% confidence interval: ±0.27	Scale AI Source details Data as of: 08/26/2026
PRBench LegalOverall Test details Test setup not documented as identicalNo direct winnerVersion: full legal setMetric: weighted rubric scoreScoring: Higher is betterClaude Opus 4.6: Reasoning: none, Attempts: pass@1 Claude Opus 4.5: Reasoning: thinking, Attempts: pass@1500 legal tasks with task-specific rubrics and o4-mini as judge.	Higher measured value52.27% ± 0.66 95% confidence interval: ±0.66	44.21% ± 0.34 95% confidence interval: ±0.34	Scale AI Source details Data as of: 08/26/2026
SecCodeBench-V2Generation with security hints Test details No documented difference in the test setupVersion: 2.2.0Metric: security-aware generation pass@1Scoring: Higher is better98 project cases. Functional tests run before dynamic exploit checks and security scoring.	70.65 %	Winner72.1 %	SecCodeBench Source details Data as of: 02/16/2026
TutorBenchOverall Test details No documented difference in the test setupVersion: 1,490-task releaseMetric: weighted rubric scoreScoring: Higher is betterSettings: Reasoning: none, Attempts: pass@11,490 tasks with weighted rubrics and Claude 4 Sonnet as judge.	Winner53.55% ± 1.01 95% confidence interval: ±1.01	49.82% ± 0.98 95% confidence interval: ±0.98	Scale AI Source details Data as of: 08/26/2026
VISTAOverall Test details No documented difference in the test setupNo clear leadVersion: 758 private prompt-image pairsMetric: rubric scoreScoring: Higher is betterSettings: Reasoning: none, Attempts: pass@1758 single-turn image-text tasks with structured yes-no rubrics.	No clear lead45.48% ± 0.44 95% confidence interval: ±0.44	45.32% ± 0.03 95% confidence interval: ±0.03	Scale AI Source details Data as of: 08/26/2026
METR Time Horizon80% time horizon Test details No documented difference in the test setupVersion: TH 1.1Metric: 80% time horizonScoring: Higher is betterClaude Opus 4.6: Snapshot: claude-opus-4-6-2026-02-05, Agent: ReAct, Harness: ReAct, Attempts: 6 independent runs per task Claude Opus 4.5: Snapshot: claude-opus-4-5-2025-11-24, Agent: ReAct, Harness: ReAct, Attempts: 6 independent runs per taskReAct agent with six independent runs per task. Time is estimated human completion duration.	Winner1.2 h (95% confidence interval 27 min to 2.8 h)	49.4 min (95% confidence interval 20.7 min to 1.7 h)	METR Source details Data as of: 05/08/2026
SecCodeBench-V2Generation without hints Test details No documented difference in the test setupVersion: 2.2.0Metric: native generation pass@1Scoring: Higher is better98 project cases. Functional tests run before dynamic exploit checks and security scoring.	51.05 %	Winner58.02 %	SecCodeBench Source details Data as of: 02/16/2026
SecCodeBench-V2Repair with security hints Test details No documented difference in the test setupVersion: 2.2.0Metric: security-aware repair pass@1Scoring: Higher is better98 project cases. Functional tests run before dynamic exploit checks and security scoring.	81.73 %	Winner82.18 %	SecCodeBench Source details Data as of: 02/16/2026
SecCodeBench-V2Repair without hints Test details No documented difference in the test setupVersion: 2.2.0Metric: native repair pass@1Scoring: Higher is better98 project cases. Functional tests run before dynamic exploit checks and security scoring.	73.1 %	Winner73.67 %	SecCodeBench Source details Data as of: 02/16/2026
SecCodeBench-V2Weighted total Test details No documented difference in the test setupVersion: 2.2.0Metric: weighted pass@1Scoring: Higher is better98 project cases. Functional tests run before dynamic exploit checks and security scoring.	64.9 %	Winner68.1 %	SecCodeBench Source details Data as of: 02/16/2026
Key to the marks:better valuebetter, but not conclusivetie

Our model tests

Our tests

SVG short test

One SVG prompt and fixed output settings. Captures are shown only with 0 reported reasoning tokens or documented lack of reasoning controls. Identical test conditions do not make the models' internal compute identical.

Shared test round: 4ae5d8eeb9a9

Claude Opus 4.6

Claude Opus 4.5

Unavailable in this round

Earlier tests

Pelican on a bicycle

A community classic for free-form SVG drawing.

Claude Opus 4.6

Reasoning (requested): high

Current attempt

Run date: Oct 1, 2026
API provider: openrouter
Requested API model: anthropic/claude-opus-4.6
API model reported in response: anthropic/claude-opus-4.6
Reasoning setting: high
Pinned endpoint: Anthropic | anthropic/claude-4.6-opus-20260205
Temperature: Provider default
Token limit including reasoning: 8,192 tokens
Test protocol: community-visual-tests-v1
Technical checks: Passed, not a quality rating
Response time: 85.1 s

Visually reviewed. This individual test does not establish overall model quality.

Claude Opus 4.5

Reasoning (requested): high

Current attempt

Run date: Oct 1, 2026
API provider: openrouter
Requested API model: anthropic/claude-opus-4.5
API model reported in response: anthropic/claude-opus-4.5
Reasoning setting: high
Pinned endpoint: Anthropic | anthropic/claude-4.5-opus-20251124
Temperature: Provider default
Token limit including reasoning: 8,192 tokens
Test protocol: community-visual-tests-v1
Technical checks: Passed, not a quality rating
Response time: 21.6 s

Visually reviewed. This individual test does not establish overall model quality.

Prompt and test conditions

Generate an SVG of a pelican riding a bicycle

Token limit including reasoning: 8,192 tokens

Task origin (Simon Willison)

How to read this comparison

Winning one benchmark is not an overall verdict

Your workload, budget, and required context length matter most. A coding benchmark says little about visual understanding or agent performance.

The standard price table uses direct developer rates. The provider comparison labels OpenRouter endpoints, regions, and special tariffs separately.

Every result links to its measurement source and retrieval date. Labels distinguish vendor reports, official benchmark leaderboards, and independent evaluations. Disputed or archived results are excluded.

More matchups

Compare Claude Opus 4.6 and Claude Opus 4.5 with other leading models using the same data and benchmark logic.

View the complete LLM comparison

Claude Opus 4.6 vs. Claude Opus 4.5

Change model selection

General specifications

API pricing compared

Shared benchmarks

Capability profile from matched benchmarks

Our tests

SVG short test

Claude Opus 4.6

Claude Opus 4.5

Pelican on a bicycle

Winning one benchmark is not an overall verdict

Related LLM comparisons