Skip to main content
BenchLM
Advertise

Model comparison

Claude Sonnet 4.5 vs Claude Sonnet 5

Data verified

Head-to-head evidence from 4 shared benchmark results across 3 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

53.61/100
Margin
11.7pts
winning →
65.32/100
2 category wins1 category wins

Public leaderboard positions: Claude Sonnet 4.5 #96 (Estimated); Claude Sonnet 5 #29 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Claude Sonnet 4.5 and Claude Sonnet 5 share 4 comparable benchmark results. 3 of 8 categories are comparable. 8 results are unique to Claude Sonnet 4.5; 32 to Claude Sonnet 5.

Updated July 23, 2026
Shared results
4
Claude Sonnet 4.5 only
8
Claude Sonnet 5 only
32
Comparable categories
3 / 8

Pick Claude Sonnet 5 if you want the stronger benchmark profile. Claude Sonnet 4.5 only becomes the better choice if knowledge is the priority or you would rather avoid the extra latency and token burn of a reasoning model.

Confidence note. This is a partial-evidence comparison with 4 shared benchmark results across 3 evidence categories; 3 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

Claude Sonnet 5 is clearly ahead on the BenchAlign aggregate, 65.32 to 53.61. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

Claude Sonnet 5's sharpest advantage is in agentic, where it averages 81.9 against 55.4. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 50% to 80.4%. Claude Sonnet 4.5 does hit back in knowledge, so the answer changes if that is the part of the workload you care about most.

Claude Sonnet 4.5 is also the more expensive model on tokens at $3.00 input / $15.00 output per 1M tokens, versus $2.00 input / $10.00 output per 1M tokens for Claude Sonnet 5. Claude Sonnet 5 is the reasoning model in the pair, while Claude Sonnet 4.5 is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. Claude Sonnet 5 gives you the larger context window at 1M, compared with 200K for Claude Sonnet 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 Sonnet 4.5 and Claude Sonnet 5
CategoryClaude Sonnet 4.5ΔClaude Sonnet 5
AgenticClaude Sonnet 4.555.4Margin→ 26.5Claude Sonnet 581.9
KnowledgeClaude Sonnet 4.583.4Margin← 26.0Claude Sonnet 557.4
CodingClaude Sonnet 4.577.2Margin← 0.5Claude Sonnet 576.7
ReasoningClaude Sonnet 4.513.6MarginNo overlapClaude Sonnet 5Not measured
MathClaude Sonnet 4.511.2MarginNo overlapClaude Sonnet 5Not measured
MultimodalClaude Sonnet 4.5Not measuredMarginNo overlapClaude Sonnet 588.3

Decisive benchmark drivers

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

More
A · Claude Sonnet 4.5B · Claude Sonnet 5

  1. Terminal-Bench 2.0

    Agentic
    Source ↗
    A 50%B 80.4%
    Winner: Claude Sonnet 5Δ 30.4
    Terminal-Bench 2.0: Claude Sonnet 4.5 scored 50%; Claude Sonnet 5 scored 80.4%. Claude Sonnet 5 wins this benchmark.

  2. OSWorld-Verified

    Agentic
    Source ↗
    A 61.4%B 81.2%
    Winner: Claude Sonnet 5Δ 19.8
    OSWorld-Verified: Claude Sonnet 4.5 scored 61.4%; Claude Sonnet 5 scored 81.2%. Claude Sonnet 5 wins this benchmark.

  3. SWE-bench Verified

    Coding
    Source ↗
    A 77.2%B 85.2%
    Winner: Claude Sonnet 5Δ 8
    SWE-bench Verified: Claude Sonnet 4.5 scored 77.2%; Claude Sonnet 5 scored 85.2%. Claude Sonnet 5 wins this benchmark.

Operational comparison

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

MetricClaude Sonnet 4.5Claude Sonnet 5Comparison
Input / output priceUSD per 1M tokensClaude Sonnet 4.5$3 input / $15 outputClaude Sonnet 5$2 input / $10 outputClaude Sonnet 5 has the lower combined listed price.
Generation speedtokens per secondClaude Sonnet 4.5Not availableClaude Sonnet 5Not availableA complete speed comparison is not available.
First-answer latencyseconds to first tokenClaude Sonnet 4.5Not availableClaude Sonnet 5Not availableA complete latency comparison is not available.
Context windowmaximum listed tokensClaude Sonnet 4.5200KClaude Sonnet 51MClaude Sonnet 5 lists the larger context window.

Benchmark Deep Dive

AgenticClaude Sonnet 5 wins
BenchmarkClaude Sonnet 4.5Claude Sonnet 5Result
Terminal-Bench 2.0Source 50%80.4%Claude Sonnet 5 leads
OSWorld-VerifiedSource 61.4%81.2%Claude Sonnet 5 leads
VITA-BenchSource 17.0%—Not comparable
Gert LabsSource 48.51%—Not comparable
JobBenchSource 27.7%—Not comparable
BrowseCompSource —84.7%Not comparable
HLE w/ toolsSource —57.4%Not comparable
GDPval-AASource —1607Not comparable
AA Agentic IndexSource —46.7%Not comparable
GDPval-AASource —55.4%Not comparable
AA BriefcaseSource —1388Not comparable
AA AutomationBenchSource —39.2%Not comparable
AA EnterpriseOps-GymSource —44.7%Not comparable
AA Harvey LABSource —90.1%Not comparable
AA Tau3 BankingSource —28.2%Not comparable
aaTerminalBench21Source —80.5%Not comparable
CodingClaude Sonnet 4.5 wins
BenchmarkClaude Sonnet 4.5Claude Sonnet 5Result
SWE-bench VerifiedSource 77.2%85.2%Claude Sonnet 5 leads
SWE-bench ProSource —63.2%Not comparable
SWE MultilingualSource —78.3%Not comparable
SWE MultimodalSource —28.1%Not comparable
Terminal-Bench 2.0Source —80.4%Not comparable
FrontierCode 1.1 MainSource —42.7%Not comparable
cursorBench32Source —61.5%Not comparable
AA Coding IndexSource —71.5%Not comparable
AA-SciCodeSource —53.6%Not comparable
Reasoning
BenchmarkClaude Sonnet 4.5Claude Sonnet 5Result
ARC-AGI-2Source 13.6%—Not comparable
AA-LCRSource —70.7%Not comparable
CritPtSource —16.9%Not comparable
KnowledgeClaude Sonnet 4.5 wins
BenchmarkClaude Sonnet 4.5Claude Sonnet 5Result
GPQASource 83.4%—Not comparable
HLESource —57.4%Not comparable
HLE w/o toolsSource —43.2%Not comparable
Artificial Analysis Intelligence IndexSource —53.4%Not comparable
AA-GPQA DiamondSource —91.1%Not comparable
AA-HLESource —39.6%Not comparable
AA-Omniscience IndexSource —15.3%Not comparable
AA-Omniscience AccuracySource —38.3%Not comparable
AA-Omniscience Hallucination RateSource —37.3%Not comparable
Math
BenchmarkClaude Sonnet 4.5Claude Sonnet 5Result
AIME 2025Source 87%—Not comparable
FrontierMath v2 (Tiers 1-3)Source 13.495%—Not comparable
FrontierMath v2 (Tier 4)Source 4.167%—Not comparable
Multimodal
BenchmarkClaude Sonnet 4.5Claude Sonnet 5Result
Design Arena WebsiteSource 12191314Claude Sonnet 5 leads
CharXivSource —88.3%Not comparable
CharXiv w/o toolsSource —77%Not comparable
AA-MMMU-ProSource —77.3%Not comparable
Frequently Asked Questions (4)

Which is better, Claude Sonnet 4.5 or Claude Sonnet 5?

Claude Sonnet 5 is ahead on BenchLM's BenchAlign leaderboard, 65.32 to 53.61. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 50% and 80.4%.

Which is better for knowledge tasks, Claude Sonnet 4.5 or Claude Sonnet 5?

Claude Sonnet 4.5 has the edge for knowledge tasks in this comparison, averaging 83.4 versus 57.4. Claude Sonnet 5 stays close enough that the answer can still flip depending on your workload.

Which is better for coding, Claude Sonnet 4.5 or Claude Sonnet 5?

Claude Sonnet 4.5 has the edge for coding in this comparison, averaging 77.2 versus 76.7. Inside this category, SWE-bench Verified is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, Claude Sonnet 4.5 or Claude Sonnet 5?

Claude Sonnet 5 has the edge for agentic tasks in this comparison, averaging 81.9 versus 55.4. Inside this category, Terminal-Bench 2.0 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 23, 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.