Compare · ModelsLive · 2 picked · head to head
Claude Sonnet 4.6 vs DeepSeek V3.2
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Claude Sonnet 4.6 wins on 16/18 benchmarks
Claude Sonnet 4.6 wins 16 of 18 shared benchmarks. Leads in speed · agentic · reasoning.
Category leads
speed·Claude Sonnet 4.6agentic·Claude Sonnet 4.6reasoning·Claude Sonnet 4.6arena·Claude Sonnet 4.6knowledge·Claude Sonnet 4.6general·Claude Sonnet 4.6math·Claude Sonnet 4.6coding·Claude Sonnet 4.6
Hype vs Reality
Attention vs performance
Claude Sonnet 4.6
#156 by perf·#12 by attention
DeepSeek V3.2
#145 by perf·no signal
Best value
DeepSeek V3.2
26.7x better value than Claude Sonnet 4.6
Claude Sonnet 4.6
5.2 pts/$
$9.00/M
DeepSeek V3.2
137.7 pts/$
$0.35/M
Vendor risk
Mixed exposure
One or more vendors flagged
Anthropic
$965.0B·Tier 1
DeepSeek
$3.4B·Tier 1
Head to head
18 benchmarks · 2 models
Claude Sonnet 4.6DeepSeek V3.2
Artificial Analysis · Agentic Index
DeepSeek V3.2 leads by +12.1
Artificial Analysis Agentic Index · a composite score measuring how well a model performs in agentic workflows · multi-step tool use, planning, error recovery, and autonomous task completion. Aggregates results from multiple agentic benchmarks including SWE-bench, tool-use tests, and planning evaluations. The canonical single-number metric for "how good is this model as an agent?"
Claude Sonnet 4.6
40.8
DeepSeek V3.2
52.9
Artificial Analysis · Coding Index
Claude Sonnet 4.6 leads by +26.3
Artificial Analysis Coding Index · a composite score that aggregates performance across multiple coding benchmarks into a single index. Tracks code generation quality, debugging ability, multi-language competence, and real-world software engineering tasks. Used by Artificial Analysis to rank model coding capability in a normalized, comparable format. Useful for developers choosing between models for coding-heavy workloads.
Claude Sonnet 4.6
63.0
DeepSeek V3.2
36.7
Artificial Analysis · Quality Index
Claude Sonnet 4.6 leads by +5.5
Claude Sonnet 4.6
47.2
DeepSeek V3.2
41.7
APEX-Agents
Claude Sonnet 4.6 leads by +21.7
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
Claude Sonnet 4.6
43.0
DeepSeek V3.2
21.3
ARC-AGI
Claude Sonnet 4.6 leads by +29.5
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Claude Sonnet 4.6
86.5
DeepSeek V3.2
57.0
ARC-AGI-2
Claude Sonnet 4.6 leads by +56.4
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
Claude Sonnet 4.6
60.4
DeepSeek V3.2
4.0
Chatbot Arena Elo · Coding
Claude Sonnet 4.6 leads by +196.6
Claude Sonnet 4.6
1521.4
DeepSeek V3.2
1324.8
Chatbot Arena Elo · Overall
Claude Sonnet 4.6 leads by +47.4
Claude Sonnet 4.6
1472.2
DeepSeek V3.2
1424.8
Chess Puzzles
Claude Sonnet 4.6 leads by +9.5
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Claude Sonnet 4.6
19.0
DeepSeek V3.2
9.5
Dtbench
Claude Sonnet 4.6 leads by +7.1
Claude Sonnet 4.6
83.1
DeepSeek V3.2
76.0
FrontierMath-2025-02-28-Private
Claude Sonnet 4.6 leads by +18.1
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
Claude Sonnet 4.6
56.8
DeepSeek V3.2
38.8
FrontierMath-Tier-4-2025-07-01-Private
Claude Sonnet 4.6 leads by +10.3
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
Claude Sonnet 4.6
13.8
DeepSeek V3.2
3.5
GPQA diamond
Claude Sonnet 4.6 leads by +5.3
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Claude Sonnet 4.6
83.2
DeepSeek V3.2
77.9
Lmca
Claude Sonnet 4.6 leads by +20.8
Claude Sonnet 4.6
54.7
DeepSeek V3.2
33.9
OTIS Mock AIME 2024-2025
DeepSeek V3.2 leads by +2.0
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Claude Sonnet 4.6
85.8
DeepSeek V3.2
87.8
Proofbench
Claude Sonnet 4.6 leads by +37.0
Claude Sonnet 4.6
45.0
DeepSeek V3.2
8.0
SimpleQA Verified
Claude Sonnet 4.6 leads by +8.0
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Claude Sonnet 4.6
35.5
DeepSeek V3.2
27.5
Terminal Bench
Claude Sonnet 4.6 leads by +13.9
Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence.
Claude Sonnet 4.6
53.4
DeepSeek V3.2
39.5
Full benchmark table
| Benchmark | Claude Sonnet 4.6 | DeepSeek V3.2 |
|---|---|---|
Artificial Analysis · Agentic Index Artificial Analysis Agentic Index · a composite score measuring how well a model performs in agentic workflows · multi-step tool use, planning, error recovery, and autonomous task completion. Aggregates results from multiple agentic benchmarks including SWE-bench, tool-use tests, and planning evaluations. The canonical single-number metric for "how good is this model as an agent?" | 40.8 | 52.9 |
Artificial Analysis · Coding Index Artificial Analysis Coding Index · a composite score that aggregates performance across multiple coding benchmarks into a single index. Tracks code generation quality, debugging ability, multi-language competence, and real-world software engineering tasks. Used by Artificial Analysis to rank model coding capability in a normalized, comparable format. Useful for developers choosing between models for coding-heavy workloads. | 63.0 | 36.7 |
Artificial Analysis · Quality Index | 47.2 | 41.7 |
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | 43.0 | 21.3 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 86.5 | 57.0 |
ARC-AGI-2 ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data. | 60.4 | 4.0 |
Chatbot Arena Elo · Coding | 1521.4 | 1324.8 |
Chatbot Arena Elo · Overall | 1472.2 | 1424.8 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 19.0 | 9.5 |
Dtbench | 83.1 | 76.0 |
FrontierMath-2025-02-28-Private FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning. | 56.8 | 38.8 |
FrontierMath-Tier-4-2025-07-01-Private FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning. | 13.8 | 3.5 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 83.2 | 77.9 |
Lmca | 54.7 | 33.9 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 85.8 | 87.8 |
Proofbench | 45.0 | 8.0 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 35.5 | 27.5 |
Terminal Bench Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence. | 53.4 | 39.5 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $3.00 | $15.00 | 1.0M tokens (~500 books) | $60.00 | |
| $0.28 | $0.42 | 164K tokens (~82 books) | $3.15 |
People also compared