LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

Claude Opus 4.6 vs DeepSeek V3.2

Updated October 21, 2026. 69 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace Claude Opus 4.6 Replace DeepSeek V3.2
Claude Opus 4.6Anthropic
Coding
66.9
Overall
31.9
Output $/M
$25.00
68
DeepSeek V3.2DeepSeek
Coding
54.4
Overall
21.5
Output $/M
$0.42
1
vs boards won of 69 shared

Decision reading

Pick Claude Opus 4.6 for coding: it scores 66.9 on the Vibe Coding Index against 54.4 for DeepSeek V3.2.

Pick DeepSeek V3.2 when output tokens cost 59.5x less, or when a solved task costs $0.99 against $1.56.

Model AClaude Opus 4.6Anthropic66.9/1002 independent benchmarks$1.56 per solved taskModel BDeepSeek V3.2DeepSeek54.4/1002 independent benchmarks$0.99 per solved task
Shared results
69
Claude Opus 4.6 only
39
DeepSeek V3.2 only
6
Like-for-like categories
8 of 9

Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.

Which one for your work

Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.

At a glance

Coding index
Claude Opus 4.6 66.9 from 2 benchmarks, DeepSeek V3.2 54.4 from 2 benchmarks.
Widest category
Agentic work, 23.4 points apart on average over 3 shared benchmarks.
Widest single gap
ARC-AGI-2: Claude Opus 4.6 69.2% against DeepSeek V3.2 4.0%, 65.1 points apart on a 0 to 100 scale.
Output price
Claude Opus 4.6 $25.00 against DeepSeek V3.2 $0.42 per million output tokens, 59.5x apart.
Context window
Claude Opus 4.6 takes 1M tokens, DeepSeek V3.2 164K.

Benchmark by benchmark

First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.

Boards won, where both have a result
CategoryClaude Opus 4.6BarsDeepSeek V3.2Ties
Coding120
Intelligence20
Agentic work30
Reasoning80
Math30
Knowledge100
Web design80
General221
All shared boards681
CodingLike-for-likeFour of these boards feed the Vibe Coding Index120
Coding benchmarks for Claude Opus 4.6 and DeepSeek V3.2
BenchmarkClaude Opus 4.6DeepSeek V3.2Leader
SWE-bench Verified75.6%70.0%Claude Opus 4.6
SWE-bench Multilingual72.0%59.0%Claude Opus 4.6
LMArena WebDev15471362Claude Opus 4.6
AA Coding IndexNot measured44.2Only DeepSeek V3.2 measured
LiveCodeBench (Vals AI)84.7%80.7%Claude Opus 4.6
SWE-bench Verified (Vals AI)78.2%67.6%Claude Opus 4.6
Vibe Code Bench (Vals AI)57.6%5.1%Claude Opus 4.6
LiveCodeBench (Artificial Analysis run)Not measured86.2%Only DeepSeek V3.2 measured
Terminal-Bench Hard (Artificial Analysis run)48.5%35.6%Claude Opus 4.6
Terminal-Bench79.8%39.6%Claude Opus 4.6
FrontierCode26.6%Not measuredOnly Claude Opus 4.6 measured
GSO-Bench41.2%Not measuredOnly Claude Opus 4.6 measured
WeirdML78.0%Not measuredOnly Claude Opus 4.6 measured
Cybench93.0%Not measuredOnly Claude Opus 4.6 measured
LMArena Text: coding15341453Claude Opus 4.6
LMArena Text: software and IT15271449Claude Opus 4.6
LMArena WebDev: React15391350Claude Opus 4.6
LMArena WebDev: HTML15591371Claude Opus 4.6
LMArena WebDev: from an image1543Not measuredOnly Claude Opus 4.6 measured
Neither model has a result on the other 19 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
Intelligence benchmarks for Claude Opus 4.6 and DeepSeek V3.2
BenchmarkClaude Opus 4.6DeepSeek V3.2Leader
AA Intelligence Index31.921.5Claude Opus 4.6
Epoch Capabilities Index155.7146.1Claude Opus 4.6
Agentic workLike-for-likePublic boards, shown as published, kept out of the index30
Agentic work benchmarks for Claude Opus 4.6 and DeepSeek V3.2
BenchmarkClaude Opus 4.6DeepSeek V3.2Leader
Design Arena: agents building web apps1206Not measuredOnly Claude Opus 4.6 measured
Design Arena: agents building full-stack apps1220Not measuredOnly Claude Opus 4.6 measured
Design Arena: agents building mobile apps1207Not measuredOnly Claude Opus 4.6 measured
Design Arena: agents building Android apps1204Not measuredOnly Claude Opus 4.6 measured
Design Arena: agents building games1214Not measuredOnly Claude Opus 4.6 measured
τ²-Bench Telecom92.1%90.6%Claude Opus 4.6
APEX-Agents46.3%21.3%Claude Opus 4.6
METR time horizons78.9%Not measuredOnly Claude Opus 4.6 measured
Remote Labor Index4.2%Not measuredOnly Claude Opus 4.6 measured
DeepResearch Bench55.3%Not measuredOnly Claude Opus 4.6 measured
Vending-Bench 2$8,018$1,034Claude Opus 4.6
Neither model has a result on the other 28 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index80
Reasoning benchmarks for Claude Opus 4.6 and DeepSeek V3.2
BenchmarkClaude Opus 4.6DeepSeek V3.2Leader
GPQA Diamond89.7%80.3%Claude Opus 4.6
MMMU Pro (Vals AI)83.9%Not measuredOnly Claude Opus 4.6 measured
GPQA Diamond (Artificial Analysis run)89.6%84.0%Claude Opus 4.6
Humanity's Last Exam (Artificial Analysis run)39.9%24.6%Claude Opus 4.6
GPQA Diamond (Epoch AI run)90.5%Not measuredOnly Claude Opus 4.6 measured
ARC-AGI94.0%57.0%Claude Opus 4.6
ARC-AGI-269.2%4.0%Claude Opus 4.6
SimpleBench67.6%Not measuredOnly Claude Opus 4.6 measured
Chess puzzles21.0%Not measuredOnly Claude Opus 4.6 measured
Mystery game puzzles25.0%Not measuredOnly Claude Opus 4.6 measured
EBR-bench12.7%Not measuredOnly Claude Opus 4.6 measured
LMArena Text: hard prompts15251434Claude Opus 4.6
LMArena Text: expert15441436Claude Opus 4.6
LMArena Text: hard prompts in English15301446Claude Opus 4.6
LMArena Vision: diagrams1340Not measuredOnly Claude Opus 4.6 measured
LMArena Vision: homework1337Not measuredOnly Claude Opus 4.6 measured
Neither model has a result on the other 5 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index30
Math benchmarks for Claude Opus 4.6 and DeepSeek V3.2
BenchmarkClaude Opus 4.6DeepSeek V3.2Leader
AIME95.6%84.6%Claude Opus 4.6
MGSM (Vals AI)Not measured86.0%Only DeepSeek V3.2 measured
AA Math IndexNot measured92.0Only DeepSeek V3.2 measured
AIME 2025 (Artificial Analysis run)Not measured92.0%Only DeepSeek V3.2 measured
OTIS Mock AIME94.4%Not measuredOnly Claude Opus 4.6 measured
FrontierMath, tiers 1 to 366.0%Not measuredOnly Claude Opus 4.6 measured
FrontierMath, tier 426.8%Not measuredOnly Claude Opus 4.6 measured
ProofBench50.0%Not measuredOnly Claude Opus 4.6 measured
LMArena Text: math15181435Claude Opus 4.6
LMArena Text: mathematical work15251440Claude Opus 4.6
Neither model has a result on the other 3 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index100
Knowledge benchmarks for Claude Opus 4.6 and DeepSeek V3.2
BenchmarkClaude Opus 4.6DeepSeek V3.2Leader
MMLU Pro89.1%84.9%Claude Opus 4.6
LegalBench (Vals AI)85.3%76.1%Claude Opus 4.6
CaseLaw (Vals AI)62.1%55.4%Claude Opus 4.6
CorpFin (Vals AI)67.0%51.0%Claude Opus 4.6
TaxEval (Vals AI)76.0%68.2%Claude Opus 4.6
MortgageTax (Vals AI)68.5%Not measuredOnly Claude Opus 4.6 measured
MedQA (Vals AI)95.4%93.9%Claude Opus 4.6
MedCode (Vals AI)49.1%Not measuredOnly Claude Opus 4.6 measured
MMLU Pro (Artificial Analysis run)Not measured86.2%Only DeepSeek V3.2 measured
SimpleQA Verified47.0%Not measuredOnly Claude Opus 4.6 measured
LMArena Text: science15191446Claude Opus 4.6
LMArena Text: medicine and healthcare15041440Claude Opus 4.6
LMArena Text: legal and government15121425Claude Opus 4.6
LMArena Text: business and finance14971410Claude Opus 4.6
Neither model has a result on the other 3 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index80
Web design benchmarks for Claude Opus 4.6 and DeepSeek V3.2
BenchmarkClaude Opus 4.6DeepSeek V3.2Leader
Design Arena: websites12911175Claude Opus 4.6
Design Arena: UI components12841149Claude Opus 4.6
Design Arena: game dev12781132Claude Opus 4.6
Design Arena: data viz12791158Claude Opus 4.6
Design Arena: 3D12741133Claude Opus 4.6
Design Arena: all code categories12891164Claude Opus 4.6
Design Arena: SVG12271032Claude Opus 4.6
Design Arena: ASCII art12541082Claude Opus 4.6
Neither model has a result on the other 1 board here
GeneralLike-for-likePublic boards, shown as published, kept out of the index221
General benchmarks for Claude Opus 4.6 and DeepSeek V3.2
BenchmarkClaude Opus 4.6DeepSeek V3.2Leader
SAGE (Vals AI)51.6%Not measuredOnly Claude Opus 4.6 measured
MedScribe (Vals AI)86.7%Not measuredOnly Claude Opus 4.6 measured
IFBench53.1%60.7%DeepSeek V3.2
Long context reasoning (AA-LCR)78.0%73.3%Claude Opus 4.6
DTBench91.2%85.6%Claude Opus 4.6
LMCA55.8%28.8%Claude Opus 4.6
CL-bench20.7%12.4%Claude Opus 4.6
Furniture assembly28.3%Not measuredOnly Claude Opus 4.6 measured
CL-bench Life17.0%7.4%Claude Opus 4.6
LMArena Text15011425Claude Opus 4.6
LMArena Text: instruction following15211413Claude Opus 4.6
LMArena Text: creative writing15051400Claude Opus 4.6
LMArena Vision1316Not measuredOnly Claude Opus 4.6 measured
LMArena Text: English15111439Claude Opus 4.6
LMArena Text: non-English14881409Claude Opus 4.6
LMArena Text: Chinese15511455Claude Opus 4.6
LMArena Text: German15001413Claude Opus 4.6
LMArena Text: Japanese14881374Claude Opus 4.6
LMArena Text: Korean14641370Claude Opus 4.6
LMArena Text: French15121429Claude Opus 4.6
LMArena Text: Spanish15101424Claude Opus 4.6
LMArena Text: Russian14971424Claude Opus 4.6
LMArena Text: Polish14991419Claude Opus 4.6
LMArena Text: multi-turn15111427Claude Opus 4.6
LMArena Text: longer queries15181428Claude Opus 4.6
LMArena Text: writing and language15021407Claude Opus 4.6
LMArena Text: entertainment and media14921395Claude Opus 4.6
LMArena Vision: English1324Not measuredOnly Claude Opus 4.6 measured
LMArena Vision: Chinese1370Not measuredOnly Claude Opus 4.6 measured
LMArena Vision: reading text1326Not measuredOnly Claude Opus 4.6 measured
LMArena Vision: humor1314Not measuredOnly Claude Opus 4.6 measured
LMArena Vision: recognizing things1274Not measuredOnly Claude Opus 4.6 measured
LMArena Vision: creative writing1331Not measuredOnly Claude Opus 4.6 measured
LMArena Document1507Not measuredOnly Claude Opus 4.6 measured
Neither model has a result on the other 12 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Claude Opus 4.6 and DeepSeek V3.2
BenchmarkClaude Opus 4.6DeepSeek V3.2Leader
Neither model has a result on 5 boards here

What each workload costs

Three fixed token mixes at each model's listed price, so per-token rates become a bill.

Chat turn

1K input + 500 output tokens

Claude Opus 4.6$0.018Fits in one request
DeepSeek V3.2$0.0005Fits in one request

DeepSeek V3.2 costs less

Repository review

50K input + 3K output tokens

Claude Opus 4.6$0.325Fits in one request
DeepSeek V3.2$0.015Fits in one request

DeepSeek V3.2 costs less

Agent loop

200K input + 10K output tokens

Claude Opus 4.6$1.25Fits in one request
DeepSeek V3.2$0.060Exceeds the context window

DeepSeek V3.2 costs less

Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.

Specification differences

The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.

Specifications of Claude Opus 4.6 and DeepSeek V3.2
SpecificationAnthropic Claude Opus 4.6DeepSeek V3.2
MakerAnthropicDeepSeek
WeightsNo public weights listedOpen weights
ReleasedFebruary 4, 2026December 1, 2025
Context window1M164K
Input $/Mtok$5.00$0.28
Output $/Mtok$25.00$0.42
Coding index66.954.4
Independent measurements behind it22
Cost per solved task$1.56$0.99
SourceOpenRouter listingOpenRouter listing

Questions

Is Claude Opus 4.6 better than DeepSeek V3.2 for coding?

Pick Claude Opus 4.6 for coding: it scores 66.9 on the Vibe Coding Index against 54.4 for DeepSeek V3.2. Pick DeepSeek V3.2 when output tokens cost 59.5x less, or when a solved task costs $0.99 against $1.56.

Which is cheaper, Claude Opus 4.6 or DeepSeek V3.2?

Claude Opus 4.6 costs $5.00 per million input tokens and $25.00 per million output tokens; DeepSeek V3.2 costs $0.28 and $0.42. Per solved SWE-bench Verified task, Claude Opus 4.6 costs $1.56 and DeepSeek V3.2 $0.99.

Which has the bigger context window, Claude Opus 4.6 or DeepSeek V3.2?

Claude Opus 4.6, with 1M tokens against 164K.

How many benchmarks is the Claude Opus 4.6 vs DeepSeek V3.2 comparison based on?

69 benchmarks have a result for both models as of October 21, 2026. Claude Opus 4.6's coding index rests on 2 independent benchmarks and DeepSeek V3.2's on 2. Where only one model has a result, the page says so instead of naming a winner.

Compare next

Embed this comparison on your site

Free under CC BY 4.0, while the source link stays visible. Preview the frame · Terms and other embeds

<iframe src="https://llmbenchmark.ai/embed/compare/claude-opus-4-6-vs-deepseek-v3-2" title="Claude Opus 4.6 vs DeepSeek V3.2, LLMBenchmark.ai" width="100%" height="360" style="border:0;max-width:100%" loading="lazy"></iframe>
<p style="font:13px/1.4 sans-serif;margin:6px 0 0">Source: <a href="https://llmbenchmark.ai/compare/claude-opus-4-6-vs-deepseek-v3-2">LLMBenchmark.ai, Claude Opus 4.6 vs DeepSeek V3.2</a></p>