Model comparison
DeepSeek V4 Pro vs Qwen2.5 32B Instruct
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 30.1 on the Noometry Index.
Last verified . 4 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 4 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 4 categories and Qwen2.5 32B Instruct in 0 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Pro leads 64.8 to 16.2.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 98.6% for DeepSeek V4 Pro and 7.4% for Qwen2.5 32B Instruct.
- DeepSeek V4 Pro is cheaper at $0.66 / $1.98 per million input/output tokens, against $0.70 / $2.80 for Qwen2.5 32B Instruct.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 131K.
Side by side
| DeepSeek V4 Pro | Qwen2.5 32B Instruct | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 54.3 | 30.1 |
| Released | 2026-04-24 | 2024-09 |
| Weights | Open | Open |
| Context window | 1M | 131K |
| Max output | 393K | 8K |
| Input $ / M tokens | $0.66 | $0.70 |
| Output $ / M tokens | $1.98 | $2.80 |
| Results tracked | 48 | 7 |
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), Qwen2.5 32B Instruct: 38.7 (#169)
| Benchmark | DeepSeek V4 Pro | Qwen2.5 32B Instruct |
|---|---|---|
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| LMArena WebDev | 1582 | — |
| SciCode | 51% | — |
| WeirdML | 66.2% | — |
| BigCodeBench Instruct | — | 45% |
| LMArena Coding | 1470 | — |
| BigCodeBench Complete | — | 52.3% |
| ALE-Bench | 1,403 | — |
Agentic & Tool Use Not comparable
DeepSeek V4 Pro: 32.8 (#58), Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek V4 Pro | Qwen2.5 32B Instruct |
|---|---|---|
| APEX-Agents | 47.3% | — |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), Qwen2.5 32B Instruct: 19.2 (#266)
| Benchmark | DeepSeek V4 Pro | Qwen2.5 32B Instruct |
|---|---|---|
| Chess Puzzles | 47% | 0% |
| Epoch Capabilities Index | 155.31 | 128.52 |
| ARC-AGI-2 | 61.3% | — |
| Kagi LLM Benchmark | 53.5% | — |
| NYT Connections (extended) | 91.3% | — |
| ARC-AGI-1 | 90.5% | — |
| CritPt | 18% | — |
| LMArena Hard Prompts | 1461 | — |
| Mystery Game Puzzles | 43% | — |
| DTBench | 93.9% | — |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
| ForecastBench | 56.1 | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), Qwen2.5 32B Instruct: 16.2 (#296)
| Benchmark | DeepSeek V4 Pro | Qwen2.5 32B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 7.4% |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| ProofBench | 50% | — |
| LMArena Math | 1455 | — |
| MATH Level 5 | — | 56.1% |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), Qwen2.5 32B Instruct: 24.9 (#266)
| Benchmark | DeepSeek V4 Pro | Qwen2.5 32B Instruct |
|---|---|---|
| GPQA Diamond | 91.7% | 46.1% |
| SimpleQA Verified | 52.9% | — |
| Vectara Hallucination Rate | 8.6% | — |
| LMArena Expert | 1464 | — |
Multilingual Not comparable
DeepSeek V4 Pro: 54.4 (#45), Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek V4 Pro | Qwen2.5 32B Instruct |
|---|---|---|
| LMArena Non-English | 1439 | — |
| LMArena Chinese | 1486 | — |
| LMArena French | 1472 | — |
| LMArena German | 1458 | — |
| LMArena Japanese | 1445 | — |
| LMArena Korean | 1447 | — |
| LMArena Russian | 1453 | — |
| LMArena Spanish | 1458 | — |
Instruction Following Not comparable
DeepSeek V4 Pro: 76.1 (#47), Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek V4 Pro | Qwen2.5 32B Instruct |
|---|---|---|
| LMArena Instruction Following | 1448 | — |
Long Context Not comparable
DeepSeek V4 Pro: 45.0 (#51), Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek V4 Pro | Qwen2.5 32B Instruct |
|---|---|---|
| CL-bench Life | 13.5% | — |
| LMArena Longer Query | 1458 | — |
Writing & Preference Not comparable
DeepSeek V4 Pro: 65.5 (#46), Qwen2.5 32B Instruct: —
| Benchmark | DeepSeek V4 Pro | Qwen2.5 32B Instruct |
|---|---|---|
| LMArena Text | 1451 | — |
| LMArena Creative Writing | 1446 | — |
| EQ-Bench Creative Writing | 1553 | — |
| EQ-Bench 4 | 1166 | — |
| LMArena Multi-Turn | 1467 | — |
Frequently asked questions
Is DeepSeek V4 Pro better than Qwen2.5 32B Instruct?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 30.1 on the Noometry Index.
Which is cheaper, DeepSeek V4 Pro or Qwen2.5 32B Instruct?
DeepSeek V4 Pro is cheaper. It lists at $0.66 per million input tokens and $1.98 per million output tokens; Qwen2.5 32B Instruct lists at $0.70 and $2.80.
Is DeepSeek V4 Pro or Qwen2.5 32B Instruct better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 38.7 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 Qwen2.5 32B Instruct share?
4 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Qwen2.5 32B Instruct has 7.