Model comparison
DeepSeek-R1 vs Qwen3 235B-A22B
Qwen3 235B-A22B is the stronger model overall, scoring 43.2 to 42.0 on the Noometry Index.
Last verified . 40 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 40 benchmarks with published results for both. DeepSeek-R1 scores higher in 4 categories and Qwen3 235B-A22B in 5 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in math, where Qwen3 235B-A22B leads 48.5 to 42.0.
- The biggest single-benchmark swing is Omni-MATH: 42.4% for DeepSeek-R1 and 71.8% for Qwen3 235B-A22B.
- DeepSeek-R1 is cheaper at $0.50 / $2.15 per million input/output tokens, against $0.70 / $2.80 for Qwen3 235B-A22B.
- DeepSeek-R1 accepts more context: 164K tokens versus 131K.
- Qwen3 235B-A22B has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-R1 | Qwen3 235B-A22B | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 42.0 | 43.2 |
| Released | 2025-01-20 | 2025-04 |
| Weights | Proprietary | Open |
| Context window | 164K | 131K |
| Max output | 64K | 16K |
| Input $ / M tokens | $0.50 | $0.70 |
| Output $ / M tokens | $2.15 | $2.80 |
| Results tracked | 52 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-R1 leads
DeepSeek-R1: 45.9 (#63), Qwen3 235B-A22B: 43.8 (#72)
| Benchmark | DeepSeek-R1 | Qwen3 235B-A22B |
|---|---|---|
| Aider Polyglot | 71.4% | 59.6% |
| SciCode | 35.7% | 42.4% |
| WeirdML | 41.6% | 41% |
| LMArena Coding | 1427 | 1445 |
| LiveBench Coding | 66.7% | — |
| ALE-Bench | 804.12 | — |
| AlgoTune | 1.7 | — |
Agentic & Tool Use Qwen3 235B-A22B leads
DeepSeek-R1: 30.4 (#68), Qwen3 235B-A22B: 33.7 (#48)
| Benchmark | DeepSeek-R1 | Qwen3 235B-A22B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 52.1% |
| DeepResearch Bench | 35.1% | — |
| BALROG | 34.9% | — |
| METR Time Horizons | 53.8% | — |
| Vending-Bench 2 | — | -11.34 |
Reasoning DeepSeek-R1 leads
DeepSeek-R1: 18.7 (#253), Qwen3 235B-A22B: 15.8 (#302)
| Benchmark | DeepSeek-R1 | Qwen3 235B-A22B |
|---|---|---|
| ARC-AGI-2 | 1.3% | 1.3% |
| SimpleBench | 40.8% | 31% |
| Kagi LLM Benchmark | 69.4% | 69.4% |
| ARC-AGI-1 | 21.2% | 11% |
| CritPt | 1.1% | 0% |
| LMArena Hard Prompts | 1415 | 1433 |
| Epoch Capabilities Index | 141.3 | 143.87 |
| ForecastBench | 60 | 59.7 |
| Chess Puzzles | — | 12% |
| LiveBench Reasoning | 83.2% | — |
| Mystery Game Puzzles | — | 9% |
| DTBench | — | 80.3% |
| LiveBench Data Analysis | 69.8% | — |
| LMCA | — | 29.3% |
| LiveBench | 71.6% | — |
Math Qwen3 235B-A22B leads
DeepSeek-R1: 42.0 (#70), Qwen3 235B-A22B: 48.5 (#51)
| Benchmark | DeepSeek-R1 | Qwen3 235B-A22B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.4% | 86.7% |
| Omni-MATH | 42.4% | 71.8% |
| LMArena Math | 1401 | 1431 |
| MATH Level 5 | 96.6% | 68.9% |
| LiveBench Math | 80.7% | — |
| FrontierMath (Feb 2025 set) | — | 8.5% |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge Qwen3 235B-A22B leads
DeepSeek-R1: 44.3 (#81), Qwen3 235B-A22B: 49.5 (#67)
| Benchmark | DeepSeek-R1 | Qwen3 235B-A22B |
|---|---|---|
| GPQA Diamond | 76.3% | 80.1% |
| MMLU-Pro | 79.3% | 84.4% |
| Confabulations | 12.7% | 15.6% |
| Vectara Hallucination Rate | 11.3% | 9.3% |
| GPQA (HELM) | 66.6% | 72.7% |
| LMArena Expert | 1394 | 1460 |
| SimpleQA Verified | — | 40.4% |
Multilingual Too close to call
DeepSeek-R1: 52.6 (#80), Qwen3 235B-A22B: 52.5 (#84)
| Benchmark | DeepSeek-R1 | Qwen3 235B-A22B |
|---|---|---|
| LMArena Non-English | 1412 | 1410 |
| LMArena Chinese | 1443 | 1481 |
| LMArena French | 1415 | 1445 |
| LMArena German | 1405 | 1432 |
| LMArena Japanese | 1394 | 1400 |
| LMArena Korean | 1359 | 1391 |
| LMArena Russian | 1423 | 1411 |
| LMArena Spanish | 1410 | 1429 |
Instruction Following Too close to call
DeepSeek-R1: 72.2 (#138), Qwen3 235B-A22B: 72.8 (#132)
| Benchmark | DeepSeek-R1 | Qwen3 235B-A22B |
|---|---|---|
| IFEval | 78.4% | 83.5% |
| LMArena Instruction Following | 1382 | 1408 |
| LiveBench Instruction Following | 80.5% | — |
Long Context Too close to call
DeepSeek-R1: 45.4 (#36), Qwen3 235B-A22B: 46.1 (#25)
| Benchmark | DeepSeek-R1 | Qwen3 235B-A22B |
|---|---|---|
| Fiction.LiveBench | 75% | 75% |
| LMArena Longer Query | 1391 | 1426 |
Writing & Preference DeepSeek-R1 leads
DeepSeek-R1: 61.5 (#84), Qwen3 235B-A22B: 59.8 (#103)
| Benchmark | DeepSeek-R1 | Qwen3 235B-A22B |
|---|---|---|
| LMArena Text | 1428 | 1419 |
| LMArena Creative Writing | 1406 | 1384 |
| Short-Story Creative Writing | 83% | 83% |
| EQ-Bench Creative Writing | 1500 | 1366 |
| WildBench | 82.8% | 86.6% |
| LMArena Multi-Turn | 1404 | 1432 |
| LiveBench Language | 48.5% | — |
Frequently asked questions
Is DeepSeek-R1 better than Qwen3 235B-A22B?
Qwen3 235B-A22B is the stronger model overall, scoring 43.2 to 42.0 on the Noometry Index.
Which is cheaper, DeepSeek-R1 or Qwen3 235B-A22B?
DeepSeek-R1 is cheaper. It lists at $0.50 per million input tokens and $2.15 per million output tokens; Qwen3 235B-A22B lists at $0.70 and $2.80.
Is DeepSeek-R1 or Qwen3 235B-A22B better for coding?
DeepSeek-R1 scores higher on coding benchmarks: 45.9 versus 43.8 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-R1 does, with 164K tokens against 131K.
How many benchmarks do DeepSeek-R1 and Qwen3 235B-A22B share?
40 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and Qwen3 235B-A22B has 49.