Model comparison
DeepSeek-V2.5 (Sep 2024) vs Qwen3.5 27B
Qwen3.5 27B is the stronger model overall, scoring 41.9 to 37.6 on the Noometry Index.
Last verified . 17 shared benchmarks.
DeepSeek
Alibaba (Qwen)Summary
- They share 17 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 0 categories and Qwen3.5 27B in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Qwen3.5 27B leads 59.3 to 49.8.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B | |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 37.6 | 41.9 |
| Released | 2024-09-06 | 2026-02-23 |
| Weights | Open | Open |
| Context window | — | 262K |
| Max output | — | 66K |
| Input $ / M tokens | — | $0.30 |
| Output $ / M tokens | — | $2.40 |
| Results tracked | 22 | 28 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.5 27B leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Qwen3.5 27B: 38.9 (#168)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| LMArena Coding | 1309 | 1427 |
| Aider Polyglot | 17.8% | — |
| LMArena WebDev | — | 1358 |
| WeirdML | — | 39.5% |
| BigCodeBench Instruct | 48.6% | — |
| BigCodeBench Complete | 53.2% | — |
| ALE-Bench | — | 349.45 |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, Qwen3.5 27B: —
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| Vending-Bench 2 | — | 201.98 |
Reasoning Qwen3.5 27B leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Qwen3.5 27B: 27.5 (#117)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1414 |
| NYT Connections (extended) | — | 47.9% |
| Thematic Generalization | — | 45.5% |
| DTBench | — | 82.4% |
| LMCA | — | 34% |
Math Qwen3.5 27B leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Qwen3.5 27B: 38.8 (#127)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| LMArena Math | 1288 | 1429 |
| MathArena Final-Answer Competitions | — | 56.7% |
Knowledge Qwen3.5 27B leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Qwen3.5 27B: 38.0 (#150)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| LMArena Expert | 1266 | 1428 |
| Vectara Hallucination Rate | — | 12.1% |
Multimodal Not comparable
DeepSeek-V2.5 (Sep 2024): —, Qwen3.5 27B: 39.4 (#59)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| LMArena Vision | — | 1241 |
Multilingual Qwen3.5 27B leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Qwen3.5 27B: 50.8 (#115)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| LMArena Non-English | 1273 | 1390 |
| LMArena Chinese | 1318 | 1478 |
| LMArena French | 1289 | 1410 |
| LMArena German | 1258 | 1393 |
| LMArena Japanese | 1228 | 1345 |
| LMArena Korean | 1209 | 1358 |
| LMArena Russian | 1289 | 1390 |
| LMArena Spanish | 1248 | 1407 |
Instruction Following Qwen3.5 27B leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Qwen3.5 27B: 73.5 (#119)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| LMArena Instruction Following | 1280 | 1393 |
Long Context Qwen3.5 27B leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Qwen3.5 27B: 43.1 (#106)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| LMArena Longer Query | 1301 | 1413 |
Writing & Preference Qwen3.5 27B leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Qwen3.5 27B: 59.3 (#111)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Qwen3.5 27B |
|---|---|---|
| LMArena Text | 1294 | 1409 |
| LMArena Creative Writing | 1285 | 1362 |
| LMArena Multi-Turn | 1297 | 1410 |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Qwen3.5 27B?
Qwen3.5 27B is the stronger model overall, scoring 41.9 to 37.6 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Qwen3.5 27B better for coding?
Qwen3.5 27B scores higher on coding benchmarks: 38.9 versus 31.7 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Qwen3.5 27B share?
17 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Qwen3.5 27B has 28.