Agentic
Like-for-like- Claude Opus 4.5
- 62.6
- Claude Sonnet 4.6
- 65.2
- Weighted basis
- 2 vs 2 rows
- Reading
- Claude Sonnet 4.6 leads
Claude Opus 4.5 vs Claude Sonnet 4.6
Updated September 3, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
Decision reading
Claude Sonnet 4.6 has the higher public score estimate, 64.57 versus 63.6, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.
14 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.
Which one for your work
Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.
Agentic work
Tool use, computer use, and multi-step task completion
Claude Sonnet 4.6
Claude Sonnet 4.6 leads on the same 2 weighted benchmark rows.
Confidence: limited
Chat turn cost
1K fresh input + 500 output tokens
Claude Sonnet 4.6
Claude Sonnet 4.6 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Show secondary and unsupported calls
Repository review cost
50K fresh input + 3K output tokens
Claude Sonnet 4.6
Claude Sonnet 4.6 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.
Confidence: limited
Long documents
Prompts that approach the documented context limit
No clear pick
The documented context windows are equal.
Confidence: documented
Cache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
Not enough matched evidence
The page does not recommend a cost winner because at least one model cannot fit the stated workload in one request. Claude Opus 4.5 does not fit this workload in one request. Claude Sonnet 4.6 does not fit this workload in one request. Claude Opus 4.5 has no published cached-input rate, so cached tokens use its listed input rate. Claude Sonnet 4.6 has no published cached-input rate, so cached tokens use its listed input rate.
Confidence: rate-fallback
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
31 Claude Opus 4.5 only
14 Shared
6 Claude Sonnet 4.6 only
- Shared results
- 14
- Claude Opus 4.5 only
- 31
- Claude Sonnet 4.6 only
- 6
- Like-for-like categories
- 2 / 8
3 categories use different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Category results, on a stated basis
Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.
Knowledge
Like-for-like- Claude Opus 4.5
- 58.1
- Claude Sonnet 4.6
- 66.0
- Weighted basis
- 4 vs 4 rows
- Reading
- Claude Sonnet 4.6 leads
Coding
Directional only- Claude Opus 4.5
- 71.7
- Claude Sonnet 4.6
- 69.1
- Weighted basis
- 2 vs 2 rows
- Reading
- Directional only
Math
Directional only- Claude Opus 4.5
- 57.5
- Claude Sonnet 4.6
- 26.4
- Weighted basis
- 4 vs 2 rows
- Reading
- Directional only
Multimodal
Directional only- Claude Opus 4.5
- 69.9
- Claude Sonnet 4.6
- 77.4
- Weighted basis
- 2 vs 1 rows
- Reading
- Directional only
Reasoning
Not comparable- Claude Opus 4.5
- 64.4
- Claude Sonnet 4.6
- Not measured
- Weighted basis
- 1 vs 0 rows
- Reading
- Not comparable
Multilingual
Not comparable- Claude Opus 4.5
- 85.7
- Claude Sonnet 4.6
- Not measured
- Weighted basis
- 1 vs 0 rows
- Reading
- Not comparable
Instruction following
Not comparable- Claude Opus 4.5
- 69.5
- Claude Sonnet 4.6
- Not measured
- Weighted basis
- 2 vs 0 rows
- Reading
- Not comparable
| Category | Claude Opus 4.5 | Claude Sonnet 4.6 | Weighted basis | Reading |
|---|---|---|---|---|
| Agentic | 62.6 | 65.2 | Like-for-like2 vs 2 rows | Claude Sonnet 4.6 leads |
| Knowledge | 58.1 | 66.0 | Like-for-like4 vs 4 rows | Claude Sonnet 4.6 leads |
| Coding | 71.7 | 69.1 | Directional only2 vs 2 rows | Directional only |
| Math | 57.5 | 26.4 | Directional only4 vs 2 rows | Directional only |
| Multimodal | 69.9 | 77.4 | Directional only2 vs 1 rows | Directional only |
| Reasoning | 64.4 | Not measured | Not comparable1 vs 0 rows | Not comparable |
| Multilingual | 85.7 | Not measured | Not comparable1 vs 0 rows | Not comparable |
| Instruction following | 69.5 | Not measured | Not comparable2 vs 0 rows | Not comparable |
Shape of the matched evidence
Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.
Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.
SuperGPQA
Knowledge
HLE
Knowledge
Claude Opus 4.5: 30.8%Claude Sonnet 4.6: 49%Normalized gap 18.2Claude Opus 4.5 source Claude Sonnet 4.6 sourceFrontierMath v2 (Tiers 1-3)
Math
MMLU-Pro
Knowledge
CharXiv
Multimodal
Claude Opus 4.5: 68.5%Claude Sonnet 4.6: 77.4%Normalized gap 8.9Claude Opus 4.5 source Claude Sonnet 4.6 source
What each workload costs
Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.
Chat turn
1K fresh input + 500 output tokens
- Claude Opus 4.5
- $0.0175
- Fits in one request
- Claude Sonnet 4.6
- $0.0105
- Fits in one request
Claude Sonnet 4.6 has the lower modeled cost
Costs use the listed standard API rates.
Repository review
50K fresh input + 3K output tokens
- Claude Opus 4.5
- $0.325
- Fits in one request
- Claude Sonnet 4.6
- $0.195
- Fits in one request
Claude Sonnet 4.6 has the lower modeled cost
Costs use the listed standard API rates.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- Claude Opus 4.5
- $1.35
- Does not fit in one request
- Cached input priced at the published list-input rate
- Claude Sonnet 4.6
- $0.81
- Does not fit in one request
- Cached input priced at the published list-input rate
Claude Opus 4.5 does not fit this workload in one request. Claude Sonnet 4.6 does not fit this workload in one request. Claude Opus 4.5 has no published cached-input rate, so cached tokens use its listed input rate. Claude Sonnet 4.6 has no published cached-input rate, so cached tokens use its listed input rate.
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationClaude Opus 4.5Claude Sonnet 4.6
Context window
Maximum documented context; output-token limits may be lower.
Claude Opus 4.5
200K
Claude Sonnet 4.6
200K
API model ID
Claude Opus 4.5
Not sourced
Claude Sonnet 4.6
Not sourced
Cached-input rate
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
Claude Opus 4.5
Not published
Claude Sonnet 4.6
Not published
Documented inputs
Claude Opus 4.5
Not sourced
Claude Sonnet 4.6
Not sourced
Documented outputs
Claude Opus 4.5
Not sourced
Claude Sonnet 4.6
Not sourced
Provider availability
Claude Opus 4.5
Not sourced
Claude Sonnet 4.6
Not sourced
Reasoning profile
Claude Opus 4.5
Non-Reasoning
Claude Sonnet 4.6
Non-Reasoning
Weight access
Claude Opus 4.5
Proprietary
Claude Sonnet 4.6
Proprietary
License
Claude Opus 4.5
Proprietary
Claude Sonnet 4.6
Proprietary
Release date
Claude Opus 4.5
2025-11-01
Claude Sonnet 4.6
2026-02-01
If you already use one of these models
- Deployment change
- Both entries list Anthropic as the provider. Confirm endpoint, model ID, limits, and feature support before switching.
- Quality signal
- Claude Sonnet 4.6 has the higher public score estimate, 64.57 versus 63.6, but the 90% score intervals overlap.
- Workload cost
- Repository review: $0.325 vs $0.195. Cache-heavy agent loop: $1.35 vs $0.81.
- Context tradeoff
- Both models list 200K.
Run the same representative tasks against both endpoints before changing production traffic.
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence51 rows
Agentic
Terminal-Bench 2.0
Claude Opus 4.559.3%SourceClaude Sonnet 4.659.1%SourceClaude Opus 4.5 leads this result
OSWorld-Verified
Claude Opus 4.566.3%SourceClaude Sonnet 4.672.1%SourceClaude Sonnet 4.6 leads this result
OSWorld
Claude Opus 4.566.3%SourceClaude Sonnet 4.6—Not directly comparable
Claw-Eval
Claude Opus 4.559.6%SourceClaude Sonnet 4.667.8%SourceClaude Sonnet 4.6 leads this result
QwenClawBench
Claude Opus 4.552.3%SourceClaude Sonnet 4.6—Not directly comparable
τ³-bench results
Claude Opus 4.570.2%SourceClaude Sonnet 4.6—Not directly comparable
VITA-Bench
Claude Opus 4.523.3%SourceClaude Sonnet 4.6—Not directly comparable
DeepPlanning
Claude Opus 4.526.4%SourceClaude Sonnet 4.6—Not directly comparable
Toolathlon
Claude Opus 4.543.5%SourceClaude Sonnet 4.6—Not directly comparable
MCP Atlas
Claude Opus 4.542.3%SourceClaude Sonnet 4.6—Not directly comparable
MCP-Tasks
Claude Opus 4.571.8%SourceClaude Sonnet 4.6—Not directly comparable
WideResearch
Claude Opus 4.576.4%SourceClaude Sonnet 4.6—Not directly comparable
CyberGym
Shared sourceClaude Opus 4.550.6%Claude Sonnet 4.665.2%Claude Sonnet 4.6 leads this result
Gert Labs
Shared sourceClaude Opus 4.564.23%Claude Sonnet 4.662.92%Claude Opus 4.5 leads this result
JobBench
Shared sourceClaude Opus 4.532.3%Claude Sonnet 4.636.9%Claude Sonnet 4.6 leads this result
OSWorld 2.0
Claude Opus 4.5—Claude Sonnet 4.68.3%SourceNot directly comparable
Coding
SWE-bench Verified
Claude Opus 4.580.9%SourceClaude Sonnet 4.679.6%SourceClaude Opus 4.5 leads this result
LiveCodeBench v6
Claude Opus 4.584.8%SourceClaude Sonnet 4.6—Not directly comparable
SWE-bench Pro
Claude Opus 4.557.1%SourceClaude Sonnet 4.6—Not directly comparable
SWE Multilingual
Claude Opus 4.577.5%SourceClaude Sonnet 4.6—Not directly comparable
NL2Repo
Claude Opus 4.543.2%SourceClaude Sonnet 4.6—Not directly comparable
SWE-Rebench
Claude Opus 4.5—Claude Sonnet 4.660.7%SourceNot directly comparable
React Native Evals
Claude Opus 4.5—Claude Sonnet 4.680.6%SourceNot directly comparable
Vibe Code Bench
Claude Opus 4.5—Claude Sonnet 4.651.48%SourceNot directly comparable
cursorBench31
Claude Opus 4.5—Claude Sonnet 4.648.8%SourceNot directly comparable
FrontierCode 1.1 Main
Claude Opus 4.5—Claude Sonnet 4.624.3%SourceNot directly comparable
Reasoning
Knowledge
GPQA
Claude Opus 4.587%SourceClaude Sonnet 4.689.9%SourceClaude Sonnet 4.6 leads this result
SuperGPQA
Shared sourceClaude Opus 4.570.6%Claude Sonnet 4.695%Claude Sonnet 4.6 leads this result
MMLU-Pro
Shared sourceClaude Opus 4.589.5%Claude Sonnet 4.679.2%Claude Opus 4.5 leads this result
MMLU-Redux
Claude Opus 4.596.6%SourceClaude Sonnet 4.6—Not directly comparable
C-Eval
Claude Opus 4.592.2%SourceClaude Sonnet 4.6—Not directly comparable
HLE
Claude Opus 4.530.8%SourceClaude Sonnet 4.649%SourceClaude Sonnet 4.6 leads this result
Math
AIME26
Claude Opus 4.595.1%SourceClaude Sonnet 4.6—Not directly comparable
HMMT Feb 2025
Claude Opus 4.592.9%SourceClaude Sonnet 4.6—Not directly comparable
HMMT Nov 2025
Claude Opus 4.593.3%SourceClaude Sonnet 4.6—Not directly comparable
HMMT Feb 2026
Claude Opus 4.585.3%SourceClaude Sonnet 4.6—Not directly comparable
MMAnswerBench
Claude Opus 4.584.0%SourceClaude Sonnet 4.6—Not directly comparable
FrontierMath v2 (Tiers 1-3)
Shared sourceClaude Opus 4.520.690%Claude Sonnet 4.632.400%Claude Sonnet 4.6 leads this result
FrontierMath v2 (Tier 4)
Shared sourceClaude Opus 4.54.167%Claude Sonnet 4.68.300%Claude Sonnet 4.6 leads this result
Multilingual
Multimodal
MMMU-Pro
Claude Opus 4.570.6%SourceClaude Sonnet 4.6—Not directly comparable
MathVision
Claude Opus 4.574.3%SourceClaude Sonnet 4.6—Not directly comparable
CharXiv
Claude Opus 4.568.5%SourceClaude Sonnet 4.677.4%SourceClaude Sonnet 4.6 leads this result
VideoMMMU
Claude Opus 4.584.4%SourceClaude Sonnet 4.6—Not directly comparable
ScreenSpot Pro
Claude Opus 4.545.7%SourceClaude Sonnet 4.6—Not directly comparable
V*
Claude Opus 4.567.0%SourceClaude Sonnet 4.6—Not directly comparable
Instruction following
Frequently asked questions
Which is better, Claude Opus 4.5 or Claude Sonnet 4.6?
Claude Sonnet 4.6 has the higher public score estimate, 64.57 versus 63.6, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.
Which is better for coding, Claude Opus 4.5 or Claude Sonnet 4.6?
The current coding averages use different weighted benchmark sets, so BenchLM does not name a winner from them. Read the shared benchmark rows directly and test the models on the same task set.
Which is better for agentic tasks, Claude Opus 4.5 or Claude Sonnet 4.6?
Claude Sonnet 4.6 leads the like-for-like agentic tasks comparison across 2 shared weighted benchmark rows.
Which costs less, Claude Opus 4.5 or Claude Sonnet 4.6?
For the stated presets, chat costs $0.0175 on Claude Opus 4.5 and $0.0105 on Claude Sonnet 4.6; repository review costs $0.325 and $0.195; the cache-heavy agent loop costs $1.35 and $0.81. Claude Opus 4.5 does not fit this workload in one request. Claude Sonnet 4.6 does not fit this workload in one request. Claude Opus 4.5 has no published cached-input rate, so cached tokens use its listed input rate. Claude Sonnet 4.6 has no published cached-input rate, so cached tokens use its listed input rate.
Which has the larger context window, Claude Opus 4.5 or Claude Sonnet 4.6?
Both models list the same context window, 200K.
Related comparisons
Last updated September 3, 2026
Watch Claude Opus 4.5 vs Claude Sonnet 4.6
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.