Model comparison
Claude Opus 4.1 vs DeepSeek-V3.2-Exp
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 41.0 on the Noometry Index.
Last verified . 29 shared benchmarks.
DeepSeekSummary
- They share 29 benchmarks with published results for both. Claude Opus 4.1 scores higher in 4 categories and DeepSeek-V3.2-Exp in 5 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3.2-Exp leads 41.7 to 22.3.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 68.9% for Claude Opus 4.1 and 87.8% for DeepSeek-V3.2-Exp.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $15 / $75 for Claude Opus 4.1.
- Claude Opus 4.1 accepts more context: 200K tokens versus 164K.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.
Side by side
| Claude Opus 4.1 | DeepSeek-V3.2-Exp | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 41.0 | 44.3 |
| Released | 2025-08-05 | 2025-09-29 |
| Weights | Proprietary | Open |
| Context window | 200K | 164K |
| Max output | 32K | 66K |
| Input $ / M tokens | $15 | $0.26 |
| Output $ / M tokens | $75 | $0.38 |
| Results tracked | 48 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Exp leads
Claude Opus 4.1: 44.4 (#73), DeepSeek-V3.2-Exp: 46.5 (#65)
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena WebDev | 1390 | 1362 |
| WeirdML | 45.9% | 39.5% |
| LMArena Coding | 1479 | 1454 |
| SWE-bench Verified | 73.3% | — |
| SWE-bench Verified (bash only) | — | 70% |
| Aider Polyglot | — | 74.2% |
| SWE-bench Multilingual | — | 59% |
| SciCode | — | 38.9% |
| ALE-Bench | 674.77 | — |
| AlgoTune | 1.34 | — |
Agentic & Tool Use Claude Opus 4.1 leads
Claude Opus 4.1: 35.0 (#41), DeepSeek-V3.2-Exp: 32.7 (#59)
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| Terminal-Bench | 38% | 39.6% |
| APEX-Agents | — | 21.3% |
| Berkeley Function Calling Leaderboard | — | 56.7% |
| GDPval | 43.6% | — |
| TheAgentCompany | — | 42.9% |
| Cybench | 42% | — |
| DeepResearch Bench | 48.3% | — |
| LMArena Search | 1148 | — |
| METR Time Horizons | 66.8% | — |
| Vending-Bench 2 | — | 1,034 |
Reasoning Claude Opus 4.1 leads
Claude Opus 4.1: 32.2 (#76), DeepSeek-V3.2-Exp: 22.1 (#208)
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| Chess Puzzles | 7% | 14% |
| LMArena Hard Prompts | 1443 | 1434 |
| DTBench | 80% | 87.7% |
| LMCA | 37.1% | 29.1% |
| Epoch Capabilities Index | 144.12 | 146.27 |
| ARC-AGI-2 | — | 4% |
| SimpleBench | 60% | — |
| Kagi LLM Benchmark | — | 52.2% |
| NYT Connections (extended) | — | 36.7% |
| ARC-AGI-1 | — | 57% |
| CritPt | — | 2.9% |
| EnigmaEval | 7.2% | — |
| Thematic Generalization | — | 65% |
| EBR-Bench | 7.9% | — |
| Mystery Game Puzzles | 21% | — |
| ForecastBench | 62 | — |
Math DeepSeek-V3.2-Exp leads
Claude Opus 4.1: 22.3 (#277), DeepSeek-V3.2-Exp: 41.7 (#87)
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 68.9% | 87.8% |
| LMArena Math | 1431 | 1435 |
| FrontierMath (Feb 2025 set) | 7.2% | 22.1% |
| FrontierMath Tier 4 (v1) | 4.2% | 2.1% |
| FrontierMath (Tiers 1-3) | 12.6% | — |
| FrontierMath Tier 4 | 2.4% | — |
| MathArena Final-Answer Competitions | — | 57.7% |
| ProofBench | — | 8% |
Knowledge DeepSeek-V3.2-Exp leads
Claude Opus 4.1: 42.0 (#101), DeepSeek-V3.2-Exp: 51.7 (#66)
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| GPQA Diamond | 77.3% | 83.4% |
| Vectara Hallucination Rate | 11.8% | 5.3% |
| LMArena Expert | 1439 | 1436 |
| Humanity's Last Exam | 11.5% | — |
| Confabulations | 17.1% | — |
Multimodal Not comparable
Claude Opus 4.1: 26.8 (#119), DeepSeek-V3.2-Exp: —
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| VPCT | 35% | — |
Multilingual Too close to call
Claude Opus 4.1: 52.0 (#95), DeepSeek-V3.2-Exp: 52.2 (#90)
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Non-English | 1405 | 1409 |
| LMArena Chinese | 1427 | 1461 |
| LMArena French | 1431 | 1433 |
| LMArena German | 1413 | 1440 |
| LMArena Japanese | 1378 | 1374 |
| LMArena Korean | 1380 | 1371 |
| LMArena Russian | 1422 | 1424 |
| LMArena Spanish | 1448 | 1440 |
Instruction Following Claude Opus 4.1 leads
Claude Opus 4.1: 75.6 (#58), DeepSeek-V3.2-Exp: 74.5 (#93)
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Instruction Following | 1435 | 1413 |
Long Context DeepSeek-V3.2-Exp leads
Claude Opus 4.1: 44.5 (#63), DeepSeek-V3.2-Exp: 47.6 (#16)
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Longer Query | 1455 | 1428 |
| Fiction.LiveBench | — | 83.3% |
| CL-bench | — | 13.2% |
| CL-bench Life | — | 9.5% |
Writing & Preference Too close to call
Claude Opus 4.1: 62.4 (#74), DeepSeek-V3.2-Exp: 62.4 (#77)
| Benchmark | Claude Opus 4.1 | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Text | 1419 | 1425 |
| LMArena Creative Writing | 1412 | 1403 |
| LMArena Multi-Turn | 1444 | 1427 |
| Short-Story Creative Writing | 84.7% | — |
| EQ-Bench Creative Writing | — | 1515 |
Frequently asked questions
Is Claude Opus 4.1 better than DeepSeek-V3.2-Exp?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 41.0 on the Noometry Index.
Which is cheaper, Claude Opus 4.1 or DeepSeek-V3.2-Exp?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Claude Opus 4.1 lists at $15 and $75.
Is Claude Opus 4.1 or DeepSeek-V3.2-Exp better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.5 versus 44.4 in the Noometry coding category.
Which has the bigger context window?
Claude Opus 4.1 does, with 200K tokens against 164K.
How many benchmarks do Claude Opus 4.1 and DeepSeek-V3.2-Exp share?
29 benchmarks have published results for both models. Claude Opus 4.1 has 48 scored results on Noometry and DeepSeek-V3.2-Exp has 49.