Model comparison
Llama 3-8B vs Qwen2.5-Coder-32B
Qwen2.5-Coder-32B is the stronger model overall, scoring 32.9 to 25.3 on the Noometry Index. Llama 3-8B costs 5.3× less per token, which makes it the better buy when Qwen2.5-Coder-32B's lead doesn't matter for your workload.
Last verified . 20 shared benchmarks.
Meta
Alibaba (Qwen)Summary
- They share 20 benchmarks with published results for both. Llama 3-8B scores higher in 1 category and Qwen2.5-Coder-32B in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Qwen2.5-Coder-32B leads 32.9 to 7.8.
- The biggest single-benchmark swing is BigCodeBench Complete: 36.9% for Llama 3-8B and 58% for Qwen2.5-Coder-32B.
- Llama 3-8B is cheaper at $0.14 / $0.14 per million input/output tokens, against $0.66 / $1 for Qwen2.5-Coder-32B.
- Qwen2.5-Coder-32B accepts more context: 33K tokens versus 8K.
Side by side
| Llama 3-8B | Qwen2.5-Coder-32B | |
|---|---|---|
| Provider | Meta | Alibaba (Qwen) |
| Noometry Index | 25.3 | 32.9 |
| Released | 2024-04-18 | 2024-09-18 |
| Weights | Open | Open |
| Context window | 8K | 33K |
| Max output | 8K | 29K |
| Input $ / M tokens | $0.14 | $0.66 |
| Output $ / M tokens | $0.14 | $1 |
| Results tracked | 34 | 31 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Llama 3-8B leads
Llama 3-8B: 30.5 (#285), Qwen2.5-Coder-32B: 22.4 (#328)
| Benchmark | Llama 3-8B | Qwen2.5-Coder-32B |
|---|---|---|
| BigCodeBench Instruct | 31.9% | 49% |
| LMArena Coding | 1152 | 1276 |
| BigCodeBench Complete | 36.9% | 58% |
| HumanEval+ | 56.7% | 87.2% |
| MBPP+ | 54.8% | 77% |
| SWE-bench Verified (bash only) | — | 9% |
| Aider Polyglot | — | 16.4% |
| LiveBench Coding | — | 56.9% |
Reasoning Qwen2.5-Coder-32B leads
Llama 3-8B: 13.8 (#321), Qwen2.5-Coder-32B: 20.4 (#220)
| Benchmark | Llama 3-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Hard Prompts | 1133 | 1251 |
| Epoch Capabilities Index | 116.33 | 119.41 |
| WinoGrande | 75.7% | 80.8% |
| Chess Puzzles | 0% | — |
| LiveBench Reasoning | — | 42.1% |
| DTBench | 43.9% | — |
| LiveBench Data Analysis | — | 49.9% |
| Adversarial NLI | 57.3% | — |
| ForecastBench | 58.6 | — |
| HellaSwag | — | 83% |
| LiveBench | — | 46.2% |
Math Qwen2.5-Coder-32B leads
Llama 3-8B: 8.3 (#317), Qwen2.5-Coder-32B: 30.8 (#203)
| Benchmark | Llama 3-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Math | 1151 | 1251 |
| OTIS Mock AIME 2024-2025 | 1.9% | — |
| LiveBench Math | — | 46.6% |
| MATH Level 5 | 6.1% | — |
| GSM8K | — | 93% |
Knowledge Qwen2.5-Coder-32B leads
Llama 3-8B: 7.8 (#303), Qwen2.5-Coder-32B: 32.9 (#200)
| Benchmark | Llama 3-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Expert | 1113 | 1221 |
| ARC (AI2) Challenge | 82.8% | 70.5% |
| MMLU | 68.8% | 79.1% |
| GPQA Diamond | 26.1% | — |
| OpenBookQA | 82.6% | — |
| TriviaQA | 67.7% | — |
Multilingual Qwen2.5-Coder-32B leads
Llama 3-8B: 31.0 (#256), Qwen2.5-Coder-32B: 38.0 (#230)
| Benchmark | Llama 3-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Non-English | 1098 | 1205 |
| LMArena Chinese | 1076 | 1222 |
| LMArena Russian | 1109 | 1228 |
| LMArena French | 1158 | — |
| LMArena German | 1104 | — |
| LMArena Japanese | 967 | — |
| LMArena Korean | 1003 | — |
| LMArena Spanish | 1172 | — |
Instruction Following Qwen2.5-Coder-32B leads
Llama 3-8B: 58.7 (#255), Qwen2.5-Coder-32B: 61.6 (#240)
| Benchmark | Llama 3-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Instruction Following | 1127 | 1223 |
| LiveBench Instruction Following | — | 58.7% |
Long Context Qwen2.5-Coder-32B leads
Llama 3-8B: 34.2 (#246), Qwen2.5-Coder-32B: 37.9 (#204)
| Benchmark | Llama 3-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Longer Query | 1128 | 1251 |
Writing & Preference Qwen2.5-Coder-32B leads
Llama 3-8B: 37.7 (#251), Qwen2.5-Coder-32B: 41.8 (#235)
| Benchmark | Llama 3-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Text | 1166 | 1230 |
| LMArena Creative Writing | 1150 | 1174 |
| LMArena Multi-Turn | 1152 | 1222 |
| LiveBench Language | — | 23.3% |
Frequently asked questions
Is Llama 3-8B better than Qwen2.5-Coder-32B?
Qwen2.5-Coder-32B is the stronger model overall, scoring 32.9 to 25.3 on the Noometry Index. Llama 3-8B costs 5.3× less per token, which makes it the better buy when Qwen2.5-Coder-32B's lead doesn't matter for your workload.
Which is cheaper, Llama 3-8B or Qwen2.5-Coder-32B?
Llama 3-8B is cheaper. It lists at $0.14 per million input tokens and $0.14 per million output tokens; Qwen2.5-Coder-32B lists at $0.66 and $1.
Is Llama 3-8B or Qwen2.5-Coder-32B better for coding?
Llama 3-8B scores higher on coding benchmarks: 30.5 versus 22.4 in the Noometry coding category.
Which has the bigger context window?
Qwen2.5-Coder-32B does, with 33K tokens against 8K.
How many benchmarks do Llama 3-8B and Qwen2.5-Coder-32B share?
20 benchmarks have published results for both models. Llama 3-8B has 34 scored results on Noometry and Qwen2.5-Coder-32B has 31.