Model comparison
Llama 2-13B vs Qwen-14B
Qwen-14B is the stronger model overall, scoring 30.8 to 29.1 on the Noometry Index.
Last verified . 18 shared benchmarks.
Meta
Alibaba (Qwen)Summary
- They share 18 benchmarks with published results for both. Llama 2-13B scores higher in 3 categories and Qwen-14B in 4 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Qwen-14B leads 18.7 to 12.3.
Side by side
| Llama 2-13B | Qwen-14B | |
|---|---|---|
| Provider | Meta | Alibaba (Qwen) |
| Noometry Index | 29.1 | 30.8 |
| Released | 2023-07-18 | 2023-09-24 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 32 | 18 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Llama 2-13B: 30.4 (#287), Qwen-14B: 30.6 (#283)
| Benchmark | Llama 2-13B | Qwen-14B |
|---|---|---|
| LMArena Coding | 1062 | 1072 |
Reasoning Qwen-14B leads
Llama 2-13B: 12.3 (#333), Qwen-14B: 18.7 (#255)
| Benchmark | Llama 2-13B | Qwen-14B |
|---|---|---|
| LMArena Hard Prompts | 1051 | 1027 |
| BIG-Bench Hard | 58.2% | 55% |
| Epoch Capabilities Index | 105.8 | 112.8 |
| LAMBADA | 76.5% | 71.1% |
| PIQA | 80.8% | 79.9% |
| Chess Puzzles | 0% | — |
| DTBench | 42.2% | — |
| HellaSwag | 80.7% | — |
| WinoGrande | 72.8% | — |
Math Too close to call
Llama 2-13B: 28.7 (#226), Qwen-14B: 28.7 (#224)
| Benchmark | Llama 2-13B | Qwen-14B |
|---|---|---|
| LMArena Math | 1065 | 1068 |
| GSM8K | 36.9% | 61.3% |
Knowledge Not comparable
Llama 2-13B: 27.5 (#244), Qwen-14B: —
| Benchmark | Llama 2-13B | Qwen-14B |
|---|---|---|
| ARC (AI2) Challenge | 60.3% | 84.4% |
| BoolQ | 82.4% | 86.2% |
| MMLU | 55.6% | 66.3% |
| LMArena Expert | 1030 | — |
| OpenBookQA | 57% | — |
| TriviaQA | 79.6% | — |
Multimodal Not comparable
Llama 2-13B: —, Qwen-14B: —
| Benchmark | Llama 2-13B | Qwen-14B |
|---|---|---|
| ScienceQA | 55.8% | — |
Multilingual Qwen-14B leads
Llama 2-13B: 26.7 (#274), Qwen-14B: 27.7 (#270)
| Benchmark | Llama 2-13B | Qwen-14B |
|---|---|---|
| LMArena Non-English | 1024 | 1042 |
| LMArena Chinese | 1001 | 1077 |
| LMArena French | 1043 | — |
| LMArena German | 1009 | — |
| LMArena Japanese | 894 | — |
| LMArena Korean | 953 | — |
| LMArena Russian | 1055 | — |
| LMArena Spanish | 1086 | — |
Instruction Following Too close to call
Llama 2-13B: 53.5 (#282), Qwen-14B: 52.6 (#284)
| Benchmark | Llama 2-13B | Qwen-14B |
|---|---|---|
| LMArena Instruction Following | 1046 | 1031 |
Long Context Llama 2-13B leads
Llama 2-13B: 32.3 (#264), Qwen-14B: 31.3 (#276)
| Benchmark | Llama 2-13B | Qwen-14B |
|---|---|---|
| LMArena Longer Query | 1064 | 1029 |
Writing & Preference Llama 2-13B leads
Llama 2-13B: 30.0 (#284), Qwen-14B: 27.8 (#294)
| Benchmark | Llama 2-13B | Qwen-14B |
|---|---|---|
| LMArena Text | 1085 | 1051 |
| LMArena Creative Writing | 1047 | 1028 |
| LMArena Multi-Turn | 1050 | 1022 |
Frequently asked questions
Is Llama 2-13B better than Qwen-14B?
Qwen-14B is the stronger model overall, scoring 30.8 to 29.1 on the Noometry Index.
Is Llama 2-13B or Qwen-14B better for coding?
They score almost the same on coding (30.4 vs 30.6); test both on your own repository before choosing.
How many benchmarks do Llama 2-13B and Qwen-14B share?
18 benchmarks have published results for both models. Llama 2-13B has 32 scored results on Noometry and Qwen-14B has 18.