Model comparison
Gemma 2 9B vs Llama 3-8B
Gemma 2 9B and Llama 3-8B score almost the same on the Noometry Index (25.9 vs 25.5), so choose on price, context window or the category you care about most.
Last verified . 24 shared benchmarks.
Google
MetaSummary
- They share 24 benchmarks with published results for both. Gemma 2 9B scores higher in 5 categories and Llama 3-8B in 3 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in multilingual, where Gemma 2 9B leads 36.6 to 30.8.
- The biggest single-benchmark swing is MATH Level 5: 21% for Gemma 2 9B and 6.1% for Llama 3-8B.
Side by side
| Gemma 2 9B | Llama 3-8B | |
|---|---|---|
| Provider | Meta | |
| Noometry Index | 25.9 | 25.5 |
| Released | 2024-06-24 | 2024-04-18 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 35 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Llama 3-8B leads
Gemma 2 9B: 29.4 (#304), Llama 3-8B: 31.0 (#289)
| Benchmark | Gemma 2 9B | Llama 3-8B |
|---|---|---|
| BigCodeBench Instruct | 34.7% | 31.9% |
| LMArena Coding | 1173 | 1152 |
| BigCodeBench Complete | 40.6% | 36.9% |
| LiveBench Coding | 22.5% | — |
| HumanEval+ | — | 56.7% |
| MBPP+ | — | 54.8% |
Reasoning Gemma 2 9B leads
Gemma 2 9B: 15.9 (#309), Llama 3-8B: 14.3 (#326)
| Benchmark | Gemma 2 9B | Llama 3-8B |
|---|---|---|
| LMArena Hard Prompts | 1171 | 1133 |
| Epoch Capabilities Index | 119.83 | 116.45 |
| Chess Puzzles | — | 0% |
| LiveBench Reasoning | 15.2% | — |
| DTBench | — | 43.9% |
| LiveBench Data Analysis | 36.4% | — |
| Adversarial NLI | — | 57.3% |
| ForecastBench | — | 58.6 |
| LiveBench | 28.7% | — |
| PIQA | 83.7% | — |
| WinoGrande | — | 75.7% |
Math Gemma 2 9B leads
Gemma 2 9B: 9.9 (#318), Llama 3-8B: 8.8 (#323)
| Benchmark | Gemma 2 9B | Llama 3-8B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 0.6% | 1.9% |
| LMArena Math | 1183 | 1151 |
| MATH Level 5 | 21% | 6.1% |
| LiveBench Math | 19.8% | — |
| GSM8K | 84.9% | — |
Knowledge Gemma 2 9B leads
Gemma 2 9B: 9.7 (#305), Llama 3-8B: 7.8 (#308)
| Benchmark | Gemma 2 9B | Llama 3-8B |
|---|---|---|
| GPQA Diamond | 27.5% | 26.1% |
| LMArena Expert | 1147 | 1113 |
| MMLU | 72.1% | 68.8% |
| ARC (AI2) Challenge | — | 82.8% |
| BoolQ | 85.7% | — |
| OpenBookQA | — | 82.6% |
| TriviaQA | — | 67.7% |
Multilingual Gemma 2 9B leads
Gemma 2 9B: 36.6 (#238), Llama 3-8B: 30.8 (#261)
| Benchmark | Gemma 2 9B | Llama 3-8B |
|---|---|---|
| LMArena Non-English | 1188 | 1098 |
| LMArena Chinese | 1185 | 1076 |
| LMArena French | 1190 | 1159 |
| LMArena German | 1186 | 1104 |
| LMArena Japanese | 1144 | 967 |
| LMArena Korean | 1137 | 1004 |
| LMArena Russian | 1200 | 1109 |
| LMArena Spanish | 1200 | 1173 |
Instruction Following Too close to call
Gemma 2 9B: 57.6 (#269), Llama 3-8B: 58.4 (#260)
| Benchmark | Gemma 2 9B | Llama 3-8B |
|---|---|---|
| LMArena Instruction Following | 1178 | 1127 |
| LiveBench Instruction Following | 52.6% | — |
Long Context Gemma 2 9B leads
Gemma 2 9B: 36.3 (#233), Llama 3-8B: 34.2 (#251)
| Benchmark | Gemma 2 9B | Llama 3-8B |
|---|---|---|
| LMArena Longer Query | 1197 | 1128 |
Writing & Preference Llama 3-8B leads
Gemma 2 9B: 32.1 (#281), Llama 3-8B: 37.5 (#256)
| Benchmark | Gemma 2 9B | Llama 3-8B |
|---|---|---|
| LMArena Text | 1207 | 1166 |
| LMArena Creative Writing | 1206 | 1150 |
| LMArena Multi-Turn | 1193 | 1152 |
| EQ-Bench Creative Writing | 841 | — |
| LiveBench Language | 25.5% | — |
Frequently asked questions
Is Gemma 2 9B better than Llama 3-8B?
Gemma 2 9B and Llama 3-8B score almost the same on the Noometry Index (25.9 vs 25.5), so choose on price, context window or the category you care about most.
Is Gemma 2 9B or Llama 3-8B better for coding?
Llama 3-8B scores higher on coding benchmarks: 31.0 versus 29.4 in the Noometry coding category.
How many benchmarks do Gemma 2 9B and Llama 3-8B share?
24 benchmarks have published results for both models. Gemma 2 9B has 35 scored results on Noometry and Llama 3-8B has 34.