Model comparison
DeepSeek V4 Pro vs Qwen3.8 Max
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 54.3 on the Noometry Index. DeepSeek V4 Pro costs 3.0× less per token, which makes it the better buy when Qwen3.8 Max's lead doesn't matter for your workload.
Last verified . 33 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 33 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 1 category and Qwen3.8 Max in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where Qwen3.8 Max leads 45.4 to 32.8.
- The biggest single-benchmark swing is FrontierMath Tier 4: 26.8% for DeepSeek V4 Pro and 46.3% for Qwen3.8 Max.
- DeepSeek V4 Pro is cheaper at $0.66 / $1.98 per million input/output tokens, against $2 / $6 for Qwen3.8 Max.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4 Pro | Qwen3.8 Max | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 54.3 | 56.8 |
| Released | 2026-04-24 | 2026-08-02 |
| Weights | Open | Proprietary |
| Context window | 1M | 1M |
| Max output | 393K | 131K |
| Input $ / M tokens | $0.66 | $2 |
| Output $ / M tokens | $1.98 | $6 |
| Results tracked | 48 | 39 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 Max leads
DeepSeek V4 Pro: 52.4 (#34), Qwen3.8 Max: 53.5 (#29)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| LMArena WebDev | 1582 | 1674 |
| SciCode | 51% | 53.2% |
| LMArena Coding | 1470 | 1502 |
| SWE-bench Verified | 77.6% | — |
| DeepSWE | — | 57.5% |
| FrontierCode | 28.6% | — |
| FrontierSWE | — | 17.8% |
| WeirdML | 66.2% | — |
| ALE-Bench | 1,403 | — |
Agentic & Tool Use Qwen3.8 Max leads
DeepSeek V4 Pro: 32.8 (#58), Qwen3.8 Max: 45.4 (#14)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| APEX-Agents | 47.3% | 63.3% |
| τ²-bench Banking | — | 55.1% |
| GDP.pdf | — | 23.2% |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), Qwen3.8 Max: 54.4 (#26)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| NYT Connections (extended) | 91.3% | 88.3% |
| CritPt | 18% | 20% |
| Chess Puzzles | 47% | 40% |
| LMArena Hard Prompts | 1461 | 1496 |
| Mystery Game Puzzles | 43% | 38% |
| DTBench | 93.9% | 92% |
| LMCA | 45.5% | 46.2% |
| Epoch Capabilities Index | 155.31 | 156.41 |
| ARC-AGI-2 | 61.3% | — |
| Kagi LLM Benchmark | 53.5% | — |
| ARC-AGI-1 | 90.5% | — |
| Surface Evolver Bench | 40% | — |
| ForecastBench | 56.1 | — |
Math Qwen3.8 Max leads
DeepSeek V4 Pro: 64.8 (#30), Qwen3.8 Max: 73.2 (#20)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| FrontierMath (Tiers 1-3) | 64.6% | 74.7% |
| FrontierMath Tier 4 | 26.8% | 46.3% |
| OTIS Mock AIME 2024-2025 | 98.6% | 100% |
| ProofBench | 50% | 58% |
| LMArena Math | 1455 | 1499 |
| MathArena Final-Answer Competitions | 76.6% | — |
Knowledge Qwen3.8 Max leads
DeepSeek V4 Pro: 59.5 (#31), Qwen3.8 Max: 61.7 (#27)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| GPQA Diamond | 91.7% | 92.7% |
| SimpleQA Verified | 52.9% | 47.3% |
| LMArena Expert | 1464 | 1507 |
| Vectara Hallucination Rate | 8.6% | — |
Multimodal Not comparable
DeepSeek V4 Pro: —, Qwen3.8 Max: 37.2 (#75)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| LMArena Vision | — | 1314 |
| Furniture Assembly | — | 20% |
Multilingual Qwen3.8 Max leads
DeepSeek V4 Pro: 54.4 (#45), Qwen3.8 Max: 56.7 (#18)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| LMArena Non-English | 1439 | 1472 |
| LMArena Chinese | 1486 | 1538 |
| LMArena French | 1472 | 1503 |
| LMArena German | 1458 | 1483 |
| LMArena Japanese | 1445 | 1467 |
| LMArena Korean | 1447 | 1461 |
| LMArena Russian | 1453 | 1481 |
| LMArena Spanish | 1458 | 1492 |
Instruction Following Qwen3.8 Max leads
DeepSeek V4 Pro: 76.1 (#47), Qwen3.8 Max: 77.6 (#17)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| LMArena Instruction Following | 1448 | 1479 |
Long Context Too close to call
DeepSeek V4 Pro: 45.0 (#51), Qwen3.8 Max: 45.6 (#31)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| LMArena Longer Query | 1458 | 1489 |
| CL-bench Life | 13.5% | — |
Writing & Preference Qwen3.8 Max leads
DeepSeek V4 Pro: 65.5 (#46), Qwen3.8 Max: 67.1 (#30)
| Benchmark | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| LMArena Text | 1451 | 1483 |
| LMArena Creative Writing | 1446 | 1479 |
| LMArena Multi-Turn | 1467 | 1489 |
| EQ-Bench Creative Writing | 1553 | — |
| EQ-Bench 4 | 1166 | — |
Frequently asked questions
Is DeepSeek V4 Pro better than Qwen3.8 Max?
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 54.3 on the Noometry Index. DeepSeek V4 Pro costs 3.0× less per token, which makes it the better buy when Qwen3.8 Max's lead doesn't matter for your workload.
Which is cheaper, DeepSeek V4 Pro or Qwen3.8 Max?
DeepSeek V4 Pro is cheaper. It lists at $0.66 per million input tokens and $1.98 per million output tokens; Qwen3.8 Max lists at $2 and $6.
Is DeepSeek V4 Pro or Qwen3.8 Max better for coding?
Qwen3.8 Max scores higher on coding benchmarks: 53.5 versus 52.4 in the Noometry coding category.
Which has the bigger context window?
Both accept 1M tokens.
How many benchmarks do DeepSeek V4 Pro and Qwen3.8 Max share?
33 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Qwen3.8 Max has 39.