Skip to main content
BenchLM
Advertise

Model comparison

Claude Opus 4.5 vs Claude Opus 4.6

Data verified

Head-to-head evidence from 29 shared benchmark results across 7 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

64.22/100
Margin
4.4pts
winning →
68.59/100
2 category wins3 category wins

Public leaderboard positions: Claude Opus 4.5 #34 (Supported); Claude Opus 4.6 #16 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Claude Opus 4.5 and Claude Opus 4.6 share 29 comparable benchmark results. 5 of 8 categories are comparable. 30 results are unique to Claude Opus 4.5; 17 to Claude Opus 4.6.

Updated July 20, 2026
Shared results
29
Claude Opus 4.5 only
30
Claude Opus 4.6 only
17
Comparable categories
5 / 8

Pick Claude Opus 4.6 if you want the stronger benchmark profile. Claude Opus 4.5 only becomes the better choice if mathematics is the priority.

Confidence note. This is a partial-evidence comparison with 29 shared benchmark results across 7 evidence categories; 5 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

Claude Opus 4.6 is clearly ahead on the BenchAlign aggregate, 68.59 to 64.22. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

Claude Opus 4.6's sharpest advantage is in knowledge, where it averages 69.1 against 58.1. The single biggest benchmark swing on the page is SuperGPQA, 70.6% to 95%. Claude Opus 4.5 does hit back in mathematics, so the answer changes if that is the part of the workload you care about most.

Claude Opus 4.6 gives you the larger context window at 1M, compared with 200K for Claude Opus 4.5.

Category breakdown

Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.

Category scores and score margins for Claude Opus 4.5 and Claude Opus 4.6
CategoryClaude Opus 4.5ΔClaude Opus 4.6
MathClaude Opus 4.557.5Margin← 21.2Claude Opus 4.636.3
KnowledgeClaude Opus 4.558.1Margin→ 11.0Claude Opus 4.669.1
AgenticClaude Opus 4.562.6Margin→ 10.4Claude Opus 4.673.0
MultimodalClaude Opus 4.569.9Margin→ 7.4Claude Opus 4.677.3
CodingClaude Opus 4.571.7Margin← 3.6Claude Opus 4.668.1
ReasoningClaude Opus 4.564.4MarginNo overlapClaude Opus 4.6Not measured
MultilingualClaude Opus 4.585.7MarginNo overlapClaude Opus 4.6Not measured
Inst. FollowingClaude Opus 4.569.5MarginNo overlapClaude Opus 4.6Not measured

Decisive benchmark drivers

The largest measured benchmark gaps in this matchup, with exact reported values.

More
A · Claude Opus 4.5B · Claude Opus 4.6

  1. SuperGPQA

    Knowledge
    Source ↗
    A 70.6%B 95%
    Winner: Claude Opus 4.6Δ 24.4
    SuperGPQA: Claude Opus 4.5 scored 70.6%; Claude Opus 4.6 scored 95%. Claude Opus 4.6 wins this benchmark.

  2. HLE

    Knowledge
    Source ↗
    A 30.8%B 53%
    Winner: Claude Opus 4.6Δ 22.2
    HLE: Claude Opus 4.5 scored 30.8%; Claude Opus 4.6 scored 53%. Claude Opus 4.6 wins this benchmark.

  3. FrontierMath v2 (Tiers 1-3)

    Math
    Source ↗
    A 20.690%B 40.700%
    Winner: Claude Opus 4.6Δ 20
    FrontierMath v2 (Tiers 1-3): Claude Opus 4.5 scored 20.690%; Claude Opus 4.6 scored 40.700%. Claude Opus 4.6 wins this benchmark.

  4. FrontierMath v2 (Tier 4)

    Math
    Source ↗
    A 4.167%B 22.900%
    Winner: Claude Opus 4.6Δ 18.7
    FrontierMath v2 (Tier 4): Claude Opus 4.5 scored 4.167%; Claude Opus 4.6 scored 22.900%. Claude Opus 4.6 wins this benchmark.

  5. MMLU-Pro

    Knowledge
    Source ↗
    A 89.5%B 82%
    Winner: Claude Opus 4.5Δ 7.5
    MMLU-Pro: Claude Opus 4.5 scored 89.5%; Claude Opus 4.6 scored 82%. Claude Opus 4.5 wins this benchmark.

Operational comparison

Runtime and commercial metrics are compared only when both models have a complete sourced value.

MetricClaude Opus 4.5Claude Opus 4.6Comparison
Input / output priceUSD per 1M tokensClaude Opus 4.5$5 input / $25 outputClaude Opus 4.6$5 input / $25 outputListed prices are equal.
Generation speedtokens per secondClaude Opus 4.546 tok/sClaude Opus 4.640 tok/sClaude Opus 4.5 has the higher measured throughput.
First-answer latencyseconds to first tokenClaude Opus 4.51.01 sClaude Opus 4.61.78 sClaude Opus 4.5 reaches the first token sooner.
Context windowmaximum listed tokensClaude Opus 4.5200KClaude Opus 4.61MClaude Opus 4.6 lists the larger context window.

Benchmark Deep Dive

AgenticClaude Opus 4.6 wins
BenchmarkClaude Opus 4.5Claude Opus 4.6Result
Terminal-Bench 2.0Source 59.3%65.4%Claude Opus 4.6 leads
OSWorld-VerifiedSource 66.3%72.7%Claude Opus 4.6 leads
OSWorldSource 66.3%—Not comparable
Claw-EvalSource 59.6%70.4%Claude Opus 4.6 leads
QwenClawBenchSource 52.3%—Not comparable
τ³-bench resultsSource 70.2%—Not comparable
VITA-BenchSource 23.3%—Not comparable
DeepPlanningSource 26.4%—Not comparable
ToolathlonSource 43.5%—Not comparable
MCP AtlasSource 42.3%—Not comparable
MCP-TasksSource 71.8%—Not comparable
WideResearchSource 76.4%—Not comparable
CyberGymSource 50.6%66.6%Claude Opus 4.6 leads
τ²-bench resultsSource 86.3%84.8%Claude Opus 4.5 leads
Gert LabsSource 64.23%61.85%Claude Opus 4.5 leads
JobBenchSource 32.3%36.7%Claude Opus 4.6 leads
BrowseCompSource —83.7%Not comparable
DeepSearchQASource —73.7%Not comparable
ResearchClawBenchSource —19.9%Not comparable
CodingClaude Opus 4.5 wins
BenchmarkClaude Opus 4.5Claude Opus 4.6Result
SWE-bench VerifiedSource 80.9%80.8%Claude Opus 4.5 leads
LiveCodeBench v6Source 84.8%—Not comparable
SWE-bench ProSource 57.1%53.4%Claude Opus 4.5 leads
SWE MultilingualSource 77.5%—Not comparable
NL2RepoSource 43.2%—Not comparable
AA-SciCodeSource 47.0%45.7%Claude Opus 4.5 leads
SWE-bench Verified*Source —75.6%Not comparable
LiveCodeBench ProSource —70.7%Not comparable
SWE-RebenchSource —65.3%Not comparable
React Native EvalsSource —84.1%Not comparable
Vibe Code BenchSource —57.57%Not comparable
FrontierCode 1.1 MainSource —26.9%Not comparable
Reasoning
BenchmarkClaude Opus 4.5Claude Opus 4.6Result
LongBench v2Source 64.4%—Not comparable
AI-NeedleSource 74%—Not comparable
AA-LCRSource 65.3%58.3%Claude Opus 4.5 leads
CritPtSource 0.3%2.8%Claude Opus 4.6 leads
KnowledgeClaude Opus 4.6 wins
BenchmarkClaude Opus 4.5Claude Opus 4.6Result
GPQASource 87%91.3%Claude Opus 4.6 leads
SuperGPQASource 70.6%95%Claude Opus 4.6 leads
MMLU-ProSource 89.5%82%Claude Opus 4.5 leads
MMLU-ReduxSource 96.6%—Not comparable
C-EvalSource 92.2%—Not comparable
HLESource 30.8%53%Claude Opus 4.6 leads
Artificial Analysis Intelligence IndexSource 34.7%37.8%Claude Opus 4.6 leads
AA-GPQA DiamondSource 81.0%84.0%Claude Opus 4.6 leads
AA-HLESource 12.9%18.6%Claude Opus 4.6 leads
AA-Omniscience IndexSource -3.9%3.5%Claude Opus 4.6 leads
AA-Omniscience AccuracySource 40.7%45.2%Claude Opus 4.6 leads
AA-Omniscience Hallucination RateSource 75.4%76.0%Claude Opus 4.5 leads
AA MMLU-ProSource 88.9%—Not comparable
GPQA-DSource —89.2%Not comparable
MMLU-Pro (Arcee)Source —89.1%Not comparable
HLE w/o toolsSource —40%Not comparable
HealthBench HardSource —14.8%Not comparable
MedXpertQA (Text)Source —52.1%Not comparable
MathClaude Opus 4.5 wins
BenchmarkClaude Opus 4.5Claude Opus 4.6Result
AIME26Source 95.1%—Not comparable
HMMT Feb 2025Source 92.9%—Not comparable
HMMT Nov 2025Source 93.3%—Not comparable
HMMT Feb 2026Source 85.3%—Not comparable
MMAnswerBenchSource 84.0%—Not comparable
FrontierMath v2 (Tiers 1-3)Source 20.690%40.700%Claude Opus 4.6 leads
FrontierMath v2 (Tier 4)Source 4.167%22.900%Claude Opus 4.6 leads
AIME25 (Arcee)Source —99.8%Not comparable
Multilingual
BenchmarkClaude Opus 4.5Claude Opus 4.6Result
MMLU-ProXSource 85.7%—Not comparable
NOVA-63Source 56.7%—Not comparable
MultimodalClaude Opus 4.6 wins
BenchmarkClaude Opus 4.5Claude Opus 4.6Result
MMMU-ProSource 70.6%77.3%Claude Opus 4.6 leads
MathVisionSource 74.3%—Not comparable
CharXivSource 68.5%—Not comparable
VideoMMMUSource 84.4%—Not comparable
ScreenSpot ProSource 45.7%83.1%Claude Opus 4.6 leads
V*Source 67.0%—Not comparable
AA-MMMU-ProSource 71.2%72.5%Claude Opus 4.6 leads
Design Arena WebsiteSource 12791328Claude Opus 4.6 leads
ERQASource —51.6%Not comparable
MedXpertQA (MM)Source —64.8%Not comparable
Inst. Following
BenchmarkClaude Opus 4.5Claude Opus 4.6Result
IFEvalSource 90.9%—Not comparable
IFBenchSource 58%—Not comparable
AA-IFBenchSource 43.0%44.6%Claude Opus 4.6 leads
Frequently Asked Questions (6)

Which is better, Claude Opus 4.5 or Claude Opus 4.6?

Claude Opus 4.6 is ahead on BenchLM's BenchAlign leaderboard, 68.59 to 64.22. The biggest single separator in this matchup is SuperGPQA, where the scores are 70.6% and 95%.

Which is better for knowledge tasks, Claude Opus 4.5 or Claude Opus 4.6?

Claude Opus 4.6 has the edge for knowledge tasks in this comparison, averaging 69.1 versus 58.1. Inside this category, SuperGPQA is the benchmark that creates the most daylight between them.

Which is better for coding, Claude Opus 4.5 or Claude Opus 4.6?

Claude Opus 4.5 has the edge for coding in this comparison, averaging 71.7 versus 68.1. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.

Which is better for math, Claude Opus 4.5 or Claude Opus 4.6?

Claude Opus 4.5 has the edge for math in this comparison, averaging 57.5 versus 36.3. Inside this category, FrontierMath v2 (Tiers 1-3) is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, Claude Opus 4.5 or Claude Opus 4.6?

Claude Opus 4.6 has the edge for agentic tasks in this comparison, averaging 73 versus 62.6. Inside this category, CyberGym is the benchmark that creates the most daylight between them.

Which is better for multimodal and grounded tasks, Claude Opus 4.5 or Claude Opus 4.6?

Claude Opus 4.6 has the edge for multimodal and grounded tasks in this comparison, averaging 77.3 versus 69.9. Inside this category, Design Arena Website is the benchmark that creates the most daylight between them.

Related Comparisons

Explore More

Anthropic Compare Pricing Methodology Find Your Best LLM Overall Rankings
Last updated: July 20, 2026

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.