Model comparison
DeepSeek V4 Pro vs Qwen3 235B-A22B
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 43.5 on the Noometry Index.
Last verified . 35 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 35 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 7 categories and Qwen3 235B-A22B in 2 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 15.7.
- The biggest single-benchmark swing is ARC-AGI-1: 90.5% for DeepSeek V4 Pro and 11% for Qwen3 235B-A22B.
- DeepSeek V4 Pro is cheaper at $0.66 / $1.98 per million input/output tokens, against $0.70 / $2.80 for Qwen3 235B-A22B.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 131K.
Side by side
| DeepSeek V4 Pro | Qwen3 235B-A22B | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 54.3 | 43.5 |
| Released | 2026-04-24 | 2025-04 |
| Weights | Open | Open |
| Context window | 1M | 131K |
| Max output | 393K | 16K |
| Input $ / M tokens | $0.66 | $0.70 |
| Output $ / M tokens | $1.98 | $2.80 |
| Results tracked | 48 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), Qwen3 235B-A22B: 44.3 (#75)
| Benchmark | DeepSeek V4 Pro | Qwen3 235B-A22B |
|---|---|---|
| SciCode | 51% | 42.4% |
| WeirdML | 66.2% | 41% |
| LMArena Coding | 1470 | 1445 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| Aider Polyglot | — | 59.6% |
| LMArena WebDev | 1582 | — |
| ALE-Bench | 1,403 | — |
Agentic & Tool Use Qwen3 235B-A22B leads
DeepSeek V4 Pro: 32.8 (#58), Qwen3 235B-A22B: 33.9 (#51)
| Benchmark | DeepSeek V4 Pro | Qwen3 235B-A22B |
|---|---|---|
| Vending-Bench 2 | 3,285 | -11.34 |
| APEX-Agents | 47.3% | — |
| Berkeley Function Calling Leaderboard | — | 52.1% |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), Qwen3 235B-A22B: 15.7 (#311)
| Benchmark | DeepSeek V4 Pro | Qwen3 235B-A22B |
|---|---|---|
| ARC-AGI-2 | 61.3% | 1.3% |
| Kagi LLM Benchmark | 53.5% | 69.4% |
| ARC-AGI-1 | 90.5% | 11% |
| CritPt | 18% | 0% |
| Chess Puzzles | 47% | 12% |
| LMArena Hard Prompts | 1461 | 1433 |
| Mystery Game Puzzles | 43% | 9% |
| DTBench | 93.9% | 80.3% |
| LMCA | 45.5% | 29.3% |
| Epoch Capabilities Index | 155.31 | 143.85 |
| ForecastBench | 56.1 | 59.7 |
| SimpleBench | — | 31% |
| NYT Connections (extended) | 91.3% | — |
| Surface Evolver Bench | 40% | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), Qwen3 235B-A22B: 50.4 (#57)
| Benchmark | DeepSeek V4 Pro | Qwen3 235B-A22B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 86.7% |
| LMArena Math | 1455 | 1432 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| ProofBench | 50% | — |
| Omni-MATH | — | 71.8% |
| MATH Level 5 | — | 68.9% |
| FrontierMath (Feb 2025 set) | — | 8.5% |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), Qwen3 235B-A22B: 49.6 (#73)
| Benchmark | DeepSeek V4 Pro | Qwen3 235B-A22B |
|---|---|---|
| GPQA Diamond | 91.7% | 80.1% |
| SimpleQA Verified | 52.9% | 40.4% |
| Vectara Hallucination Rate | 8.6% | 9.3% |
| LMArena Expert | 1464 | 1463 |
| MMLU-Pro | — | 84.4% |
| Confabulations | — | 15.6% |
| GPQA (HELM) | — | 72.7% |
Multilingual DeepSeek V4 Pro leads
DeepSeek V4 Pro: 54.4 (#45), Qwen3 235B-A22B: 52.3 (#89)
| Benchmark | DeepSeek V4 Pro | Qwen3 235B-A22B |
|---|---|---|
| LMArena Non-English | 1439 | 1409 |
| LMArena Chinese | 1486 | 1481 |
| LMArena French | 1472 | 1445 |
| LMArena German | 1458 | 1433 |
| LMArena Japanese | 1445 | 1399 |
| LMArena Korean | 1447 | 1391 |
| LMArena Russian | 1453 | 1411 |
| LMArena Spanish | 1458 | 1430 |
Instruction Following DeepSeek V4 Pro leads
DeepSeek V4 Pro: 76.1 (#47), Qwen3 235B-A22B: 72.6 (#136)
| Benchmark | DeepSeek V4 Pro | Qwen3 235B-A22B |
|---|---|---|
| LMArena Instruction Following | 1448 | 1408 |
| IFEval | — | 83.5% |
Long Context Qwen3 235B-A22B leads
DeepSeek V4 Pro: 45.0 (#51), Qwen3 235B-A22B: 46.1 (#26)
| Benchmark | DeepSeek V4 Pro | Qwen3 235B-A22B |
|---|---|---|
| LMArena Longer Query | 1458 | 1426 |
| Fiction.LiveBench | — | 75% |
| CL-bench Life | 13.5% | — |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Pro: 65.5 (#46), Qwen3 235B-A22B: 59.6 (#108)
| Benchmark | DeepSeek V4 Pro | Qwen3 235B-A22B |
|---|---|---|
| LMArena Text | 1451 | 1419 |
| LMArena Creative Writing | 1446 | 1384 |
| EQ-Bench Creative Writing | 1553 | 1366 |
| LMArena Multi-Turn | 1467 | 1432 |
| Short-Story Creative Writing | — | 83% |
| WildBench | — | 86.6% |
| EQ-Bench 4 | 1166 | — |
Frequently asked questions
Is DeepSeek V4 Pro better than Qwen3 235B-A22B?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 43.5 on the Noometry Index.
Which is cheaper, DeepSeek V4 Pro or Qwen3 235B-A22B?
DeepSeek V4 Pro is cheaper. It lists at $0.66 per million input tokens and $1.98 per million output tokens; Qwen3 235B-A22B lists at $0.70 and $2.80.
Is DeepSeek V4 Pro or Qwen3 235B-A22B better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 44.3 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Pro does, with 1M tokens against 131K.
How many benchmarks do DeepSeek V4 Pro and Qwen3 235B-A22B share?
35 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Qwen3 235B-A22B has 49.