Model comparison
DeepSeek-R1 vs DeepSeek-V3.2-Exp
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 42.3 on the Noometry Index.
Last verified . 30 shared benchmarks.
DeepSeekSummary
- They share 30 benchmarks with published results for both. DeepSeek-R1 scores higher in 2 categories and DeepSeek-V3.2-Exp in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3.2-Exp leads 51.7 to 44.5.
- The biggest single-benchmark swing is ARC-AGI-1: 21.2% for DeepSeek-R1 and 57% for DeepSeek-V3.2-Exp.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $0.50 / $2.15 for DeepSeek-R1.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-R1 | DeepSeek-V3.2-Exp | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 42.3 | 44.3 |
| Released | 2025-01-20 | 2025-09-29 |
| Weights | Proprietary | Open |
| Context window | 164K | 164K |
| Max output | 64K | 66K |
| Input $ / M tokens | $0.50 | $0.26 |
| Output $ / M tokens | $2.15 | $0.38 |
| Results tracked | 52 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
DeepSeek-R1: 46.3 (#68), DeepSeek-V3.2-Exp: 46.5 (#65)
| Benchmark | DeepSeek-R1 | DeepSeek-V3.2-Exp |
|---|---|---|
| Aider Polyglot | 71.4% | 74.2% |
| SciCode | 35.7% | 38.9% |
| WeirdML | 41.6% | 39.5% |
| LMArena Coding | 1427 | 1454 |
| SWE-bench Verified (bash only) | — | 70% |
| LMArena WebDev | — | 1362 |
| SWE-bench Multilingual | — | 59% |
| LiveBench Coding | 66.7% | — |
| ALE-Bench | 804.12 | — |
| AlgoTune | 1.7 | — |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-R1: 30.7 (#75), DeepSeek-V3.2-Exp: 32.7 (#59)
| Benchmark | DeepSeek-R1 | DeepSeek-V3.2-Exp |
|---|---|---|
| Terminal-Bench | — | 39.6% |
| APEX-Agents | — | 21.3% |
| Berkeley Function Calling Leaderboard | — | 56.7% |
| TheAgentCompany | — | 42.9% |
| DeepResearch Bench | 35.1% | — |
| BALROG | 34.9% | — |
| METR Time Horizons | 53.8% | — |
| Vending-Bench 2 | — | 1,034 |
Reasoning DeepSeek-V3.2-Exp leads
DeepSeek-R1: 18.6 (#278), DeepSeek-V3.2-Exp: 22.1 (#208)
| Benchmark | DeepSeek-R1 | DeepSeek-V3.2-Exp |
|---|---|---|
| ARC-AGI-2 | 1.3% | 4% |
| Kagi LLM Benchmark | 69.4% | 52.2% |
| ARC-AGI-1 | 21.2% | 57% |
| CritPt | 1.1% | 2.9% |
| LMArena Hard Prompts | 1416 | 1434 |
| Epoch Capabilities Index | 141.29 | 146.27 |
| SimpleBench | 40.8% | — |
| NYT Connections (extended) | — | 36.7% |
| Chess Puzzles | — | 14% |
| Thematic Generalization | — | 65% |
| LiveBench Reasoning | 83.2% | — |
| DTBench | — | 87.7% |
| LiveBench Data Analysis | 69.8% | — |
| LMCA | — | 29.1% |
| ForecastBench | 60 | — |
| LiveBench | 71.6% | — |
Math DeepSeek-R1 leads
DeepSeek-R1: 43.8 (#79), DeepSeek-V3.2-Exp: 41.7 (#87)
| Benchmark | DeepSeek-R1 | DeepSeek-V3.2-Exp |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.4% | 87.8% |
| LMArena Math | 1400 | 1435 |
| MathArena Final-Answer Competitions | — | 57.7% |
| ProofBench | — | 8% |
| Omni-MATH | 42.4% | — |
| LiveBench Math | 80.7% | — |
| MATH Level 5 | 96.6% | — |
| FrontierMath (Feb 2025 set) | — | 22.1% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge DeepSeek-V3.2-Exp leads
DeepSeek-R1: 44.5 (#87), DeepSeek-V3.2-Exp: 51.7 (#66)
| Benchmark | DeepSeek-R1 | DeepSeek-V3.2-Exp |
|---|---|---|
| GPQA Diamond | 76.3% | 83.4% |
| Vectara Hallucination Rate | 11.3% | 5.3% |
| LMArena Expert | 1394 | 1436 |
| MMLU-Pro | 79.3% | — |
| Confabulations | 12.7% | — |
| GPQA (HELM) | 66.6% | — |
Multilingual Too close to call
DeepSeek-R1: 52.4 (#85), DeepSeek-V3.2-Exp: 52.2 (#90)
| Benchmark | DeepSeek-R1 | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Non-English | 1412 | 1409 |
| LMArena Chinese | 1442 | 1461 |
| LMArena French | 1417 | 1433 |
| LMArena German | 1404 | 1440 |
| LMArena Japanese | 1391 | 1374 |
| LMArena Korean | 1360 | 1371 |
| LMArena Russian | 1423 | 1424 |
| LMArena Spanish | 1411 | 1440 |
Instruction Following DeepSeek-V3.2-Exp leads
DeepSeek-R1: 72.0 (#143), DeepSeek-V3.2-Exp: 74.5 (#93)
| Benchmark | DeepSeek-R1 | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Instruction Following | 1382 | 1413 |
| LiveBench Instruction Following | 80.5% | — |
| IFEval | 78.4% | — |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-R1: 45.4 (#36), DeepSeek-V3.2-Exp: 47.6 (#16)
| Benchmark | DeepSeek-R1 | DeepSeek-V3.2-Exp |
|---|---|---|
| Fiction.LiveBench | 75% | 83.3% |
| LMArena Longer Query | 1391 | 1428 |
| CL-bench | — | 13.2% |
| CL-bench Life | — | 9.5% |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-R1: 61.4 (#88), DeepSeek-V3.2-Exp: 62.4 (#77)
| Benchmark | DeepSeek-R1 | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Text | 1428 | 1425 |
| LMArena Creative Writing | 1405 | 1403 |
| EQ-Bench Creative Writing | 1500 | 1515 |
| LMArena Multi-Turn | 1405 | 1427 |
| Short-Story Creative Writing | 83% | — |
| WildBench | 82.8% | — |
| LiveBench Language | 48.5% | — |
Frequently asked questions
Is DeepSeek-R1 better than DeepSeek-V3.2-Exp?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 42.3 on the Noometry Index.
Which is cheaper, DeepSeek-R1 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; DeepSeek-R1 lists at $0.50 and $2.15.
Is DeepSeek-R1 or DeepSeek-V3.2-Exp better for coding?
They score almost the same on coding (46.3 vs 46.5); test both on your own repository before choosing.
Which has the bigger context window?
Both accept 164K tokens.
How many benchmarks do DeepSeek-R1 and DeepSeek-V3.2-Exp share?
30 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and DeepSeek-V3.2-Exp has 49.