Model comparison
Granite 3.0 2b Instruct vs Qwen2.5-Coder-32B
Qwen2.5-Coder-32B is the stronger model overall, scoring 33.4 to 30.8 on the Noometry Index.
Last verified . 13 shared benchmarks.
Alibaba (Qwen)Summary
- They share 13 benchmarks with published results for both. Granite 3.0 2b Instruct scores higher in 1 category and Qwen2.5-Coder-32B in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Qwen2.5-Coder-32B leads 41.6 to 29.6.
- The biggest single-benchmark swing is BigCodeBench Instruct: 20.5% for Granite 3.0 2b Instruct and 49% for Qwen2.5-Coder-32B.
Side by side
| Granite 3.0 2b Instruct | Qwen2.5-Coder-32B | |
|---|---|---|
| Provider | IBM | Alibaba (Qwen) |
| Noometry Index | 30.8 | 33.4 |
| Released | — | 2024-09-18 |
| Weights | Open | Open |
| Context window | — | 33K |
| Max output | — | 29K |
| Input $ / M tokens | — | $0.66 |
| Output $ / M tokens | — | $1 |
| Results tracked | 13 | 31 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Granite 3.0 2b Instruct leads
Granite 3.0 2b Instruct: 28.3 (#316), Qwen2.5-Coder-32B: 22.6 (#333)
| Benchmark | Granite 3.0 2b Instruct | Qwen2.5-Coder-32B |
|---|---|---|
| BigCodeBench Instruct | 20.5% | 49% |
| LMArena Coding | 1090 | 1276 |
| SWE-bench Verified (bash only) | — | 9% |
| Aider Polyglot | — | 16.4% |
| LiveBench Coding | — | 56.9% |
| BigCodeBench Complete | — | 58% |
| HumanEval+ | — | 87.2% |
| MBPP+ | — | 77% |
Reasoning Too close to call
Granite 3.0 2b Instruct: 20.5 (#235), Qwen2.5-Coder-32B: 21.2 (#225)
| Benchmark | Granite 3.0 2b Instruct | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Hard Prompts | 1073 | 1251 |
| LiveBench Reasoning | — | 42.1% |
| LiveBench Data Analysis | — | 49.9% |
| Epoch Capabilities Index | — | 119.49 |
| HellaSwag | — | 83% |
| LiveBench | — | 46.2% |
| WinoGrande | — | 80.8% |
Math Qwen2.5-Coder-32B leads
Granite 3.0 2b Instruct: 32.2 (#217), Qwen2.5-Coder-32B: 33.3 (#204)
| Benchmark | Granite 3.0 2b Instruct | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Math | 1117 | 1251 |
| LiveBench Math | — | 46.6% |
| GSM8K | — | 93% |
Knowledge Qwen2.5-Coder-32B leads
Granite 3.0 2b Instruct: 29.0 (#241), Qwen2.5-Coder-32B: 33.4 (#203)
| Benchmark | Granite 3.0 2b Instruct | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Expert | 1064 | 1221 |
| ARC (AI2) Challenge | — | 70.5% |
| MMLU | — | 79.1% |
Multilingual Qwen2.5-Coder-32B leads
Granite 3.0 2b Instruct: 27.0 (#278), Qwen2.5-Coder-32B: 37.8 (#235)
| Benchmark | Granite 3.0 2b Instruct | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Non-English | 1033 | 1205 |
| LMArena Chinese | 1070 | 1222 |
| LMArena Russian | 1045 | 1228 |
Instruction Following Qwen2.5-Coder-32B leads
Granite 3.0 2b Instruct: 53.9 (#284), Qwen2.5-Coder-32B: 61.4 (#245)
| Benchmark | Granite 3.0 2b Instruct | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Instruction Following | 1056 | 1223 |
| LiveBench Instruction Following | — | 58.7% |
Long Context Qwen2.5-Coder-32B leads
Granite 3.0 2b Instruct: 32.5 (#268), Qwen2.5-Coder-32B: 38.0 (#208)
| Benchmark | Granite 3.0 2b Instruct | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Longer Query | 1070 | 1251 |
Writing & Preference Qwen2.5-Coder-32B leads
Granite 3.0 2b Instruct: 29.6 (#292), Qwen2.5-Coder-32B: 41.6 (#240)
| Benchmark | Granite 3.0 2b Instruct | Qwen2.5-Coder-32B |
|---|---|---|
| LMArena Text | 1080 | 1230 |
| LMArena Creative Writing | 1046 | 1174 |
| LMArena Multi-Turn | 1053 | 1222 |
| LiveBench Language | — | 23.3% |
Frequently asked questions
Is Granite 3.0 2b Instruct better than Qwen2.5-Coder-32B?
Qwen2.5-Coder-32B is the stronger model overall, scoring 33.4 to 30.8 on the Noometry Index.
Is Granite 3.0 2b Instruct or Qwen2.5-Coder-32B better for coding?
Granite 3.0 2b Instruct scores higher on coding benchmarks: 28.3 versus 22.6 in the Noometry coding category.
How many benchmarks do Granite 3.0 2b Instruct and Qwen2.5-Coder-32B share?
13 benchmarks have published results for both models. Granite 3.0 2b Instruct has 13 scored results on Noometry and Qwen2.5-Coder-32B has 31.