Model comparison
Llama 3.1-8B vs Qwen2.5-Coder-32B
Qwen2.5-Coder-32B is the stronger model overall, scoring 33.4 to 23.0 on the Noometry Index. Llama 3.1-8B costs 13× less per token, which makes it the better buy when Qwen2.5-Coder-32B's lead doesn't matter for your workload.
Last verified . 19 shared benchmarks.
Meta
Alibaba (Qwen)Summary
- They share 19 benchmarks with published results for both. Llama 3.1-8B scores higher in 0 categories and Qwen2.5-Coder-32B in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Qwen2.5-Coder-32B leads 33.4 to 8.0.
- The biggest single-benchmark swing is BigCodeBench Complete: 40.5% for Llama 3.1-8B and 58% for Qwen2.5-Coder-32B.
- Llama 3.1-8B is cheaper at $0.05 / $0.08 per million input/output tokens, against $0.66 / $1 for Qwen2.5-Coder-32B.
- Llama 3.1-8B accepts more context: 128K tokens versus 33K.
Side by side
| Llama 3.1-8B | Qwen2.5-Coder-32B | |
|---|---|---|
| Provider | Meta | Alibaba (Qwen) |
| Noometry Index | 23.0 | 33.4 |
| Released | 2024-07-23 | 2024-09-18 |
| Weights | Open | Open |
| Context window | 128K | 33K |
| Max output | 4K | 29K |
| Input $ / M tokens | $0.05 | $0.66 |
| Output $ / M tokens | $0.08 | $1 |
| Results tracked | 43 | 31 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen2.5-Coder-32B leads
Llama 3.1-8B: 20.2 (#340), Qwen2.5-Coder-32B: 22.6 (#333)
| Benchmark | Llama 3.1-8B | Qwen2.5-Coder-32B |
|---|---|---|
| BigCodeBench Instruct | 32.8% | 49% |
| LMArena Coding | 1195 | 1276 |
| BigCodeBench Complete | 40.5% | 58% |
| HumanEval+ | 62.8% | 87.2% |
| MBPP+ | 55.6% | 77% |
| SWE-bench Verified (bash only) | — | 9% |
| Aider Polyglot | — | 16.4% |
| SciCode | 13.2% | — |
| WeirdML | 1.7% | — |
| LiveBench Coding | — | 56.9% |
Agentic & Tool Use Not comparable
Llama 3.1-8B: 22.5 (#131), Qwen2.5-Coder-32B: —
| Benchmark | Llama 3.1-8B | Qwen2.5-Coder-32B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 25.8% | — |
| BALROG | 15.1% | — |
Reasoning Qwen2.5-Coder-32B leads
Llama 3.1-8B: 14.9 (#321), Qwen2.5-Coder-32B: 21.2 (#225)
| Benchmark | Llama 3.1-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Hard Prompts | 1175 | 1251 |
| Epoch Capabilities Index | 116.57 | 119.49 |
| CritPt | 0% | — |
| Chess Puzzles | 0% | — |
| LiveBench Reasoning | — | 42.1% |
| DTBench | 50.9% | — |
| LiveBench Data Analysis | — | 49.9% |
| LMCA | 5.4% | — |
| HellaSwag | — | 83% |
| LiveBench | — | 46.2% |
| PIQA | 81.2% | — |
| WinoGrande | — | 80.8% |
Math Qwen2.5-Coder-32B leads
Llama 3.1-8B: 10.2 (#317), Qwen2.5-Coder-32B: 33.3 (#204)
| Benchmark | Llama 3.1-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Math | 1179 | 1251 |
| GSM8K | 82.4% | 93% |
| OTIS Mock AIME 2024-2025 | 1.7% | — |
| Omni-MATH | 13.7% | — |
| LiveBench Math | — | 46.6% |
| MATH Level 5 | 22.9% | — |
Knowledge Qwen2.5-Coder-32B leads
Llama 3.1-8B: 8.0 (#307), Qwen2.5-Coder-32B: 33.4 (#203)
| Benchmark | Llama 3.1-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Expert | 1144 | 1221 |
| MMLU | 56.1% | 79.1% |
| GPQA Diamond | 27% | — |
| MMLU-Pro | 40.6% | — |
| GPQA (HELM) | 24.7% | — |
| ARC (AI2) Challenge | — | 70.5% |
| BoolQ | 82.8% | — |
Multilingual Qwen2.5-Coder-32B leads
Llama 3.1-8B: 34.0 (#249), Qwen2.5-Coder-32B: 37.8 (#235)
| Benchmark | Llama 3.1-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Non-English | 1148 | 1205 |
| LMArena Chinese | 1151 | 1222 |
| LMArena Russian | 1158 | 1228 |
| LMArena French | 1177 | — |
| LMArena German | 1144 | — |
| LMArena Japanese | 1061 | — |
| LMArena Korean | 1053 | — |
| LMArena Spanish | 1169 | — |
Instruction Following Qwen2.5-Coder-32B leads
Llama 3.1-8B: 58.9 (#258), Qwen2.5-Coder-32B: 61.4 (#245)
| Benchmark | Llama 3.1-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Instruction Following | 1159 | 1223 |
| LiveBench Instruction Following | — | 58.7% |
| IFEval | 74.3% | — |
Long Context Qwen2.5-Coder-32B leads
Llama 3.1-8B: 35.8 (#238), Qwen2.5-Coder-32B: 38.0 (#208)
| Benchmark | Llama 3.1-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Longer Query | 1182 | 1251 |
Writing & Preference Qwen2.5-Coder-32B leads
Llama 3.1-8B: 29.7 (#290), Qwen2.5-Coder-32B: 41.6 (#240)
| Benchmark | Llama 3.1-8B | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Text | 1187 | 1230 |
| LMArena Creative Writing | 1154 | 1174 |
| LMArena Multi-Turn | 1172 | 1222 |
| EQ-Bench Creative Writing | 713 | — |
| WildBench | 68.7% | — |
| LiveBench Language | — | 23.3% |
Frequently asked questions
Is Llama 3.1-8B better than Qwen2.5-Coder-32B?
Qwen2.5-Coder-32B is the stronger model overall, scoring 33.4 to 23.0 on the Noometry Index. Llama 3.1-8B costs 13× 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 3.1-8B or Qwen2.5-Coder-32B?
Llama 3.1-8B is cheaper. It lists at $0.05 per million input tokens and $0.08 per million output tokens; Qwen2.5-Coder-32B lists at $0.66 and $1.
Is Llama 3.1-8B or Qwen2.5-Coder-32B better for coding?
Qwen2.5-Coder-32B scores higher on coding benchmarks: 22.6 versus 20.2 in the Noometry coding category.
Which has the bigger context window?
Llama 3.1-8B does, with 128K tokens against 33K.
How many benchmarks do Llama 3.1-8B and Qwen2.5-Coder-32B share?
19 benchmarks have published results for both models. Llama 3.1-8B has 43 scored results on Noometry and Qwen2.5-Coder-32B has 31.