Model comparison
Llama 2-70B vs phi-3-medium 14B
phi-3-medium 14B is the stronger model overall, scoring 29.7 to 24.4 on the Noometry Index.
Last verified . 10 shared benchmarks.
Meta
MicrosoftSummary
- They share 10 benchmarks with published results for both. Llama 2-70B scores higher in 0 categories and phi-3-medium 14B in 3 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in math, where phi-3-medium 14B leads 27.3 to 8.1.
- The biggest single-benchmark swing is MATH Level 5: 3.3% for Llama 2-70B and 17.6% for phi-3-medium 14B.
Side by side
| Llama 2-70B | phi-3-medium 14B | |
|---|---|---|
| Provider | Meta | Microsoft |
| Noometry Index | 24.4 | 29.7 |
| Released | 2023-07-18 | 2024-04-23 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 35 | 13 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding phi-3-medium 14B leads
Llama 2-70B: 31.4 (#286), phi-3-medium 14B: 36.8 (#201)
| Benchmark | Llama 2-70B | phi-3-medium 14B |
|---|---|---|
| BigCodeBench Instruct | — | 37.6% |
| LMArena Coding | 1079 | — |
| BigCodeBench Complete | — | 48.7% |
Reasoning Not comparable
Llama 2-70B: 14.4 (#325), phi-3-medium 14B: —
| Benchmark | Llama 2-70B | phi-3-medium 14B |
|---|---|---|
| BIG-Bench Hard | 64.9% | 81.4% |
| Epoch Capabilities Index | 113.79 | 121.23 |
| HellaSwag | 85.3% | 82.4% |
| WinoGrande | 80.2% | 81.5% |
| LMArena Hard Prompts | 1073 | — |
| DTBench | 41.6% | — |
| Adversarial NLI | — | 55.8% |
| CommonsenseQA 2.0 | 50% | — |
| ForecastBench | 51.4 | — |
| LAMBADA | 78.9% | — |
| PIQA | 82.8% | — |
Math phi-3-medium 14B leads
Llama 2-70B: 8.1 (#326), phi-3-medium 14B: 27.3 (#250)
| Benchmark | Llama 2-70B | phi-3-medium 14B |
|---|---|---|
| MATH Level 5 | 3.3% | 17.6% |
| OTIS Mock AIME 2024-2025 | 0% | — |
| LMArena Math | 1091 | — |
| GSM8K | 69.6% | — |
Knowledge phi-3-medium 14B leads
Llama 2-70B: 7.4 (#310), phi-3-medium 14B: 9.1 (#306)
| Benchmark | Llama 2-70B | phi-3-medium 14B |
|---|---|---|
| GPQA Diamond | 26.3% | 27.6% |
| ARC (AI2) Challenge | 78.3% | 91.6% |
| MMLU | 69.9% | 78% |
| OpenBookQA | 60.2% | 87.4% |
| TriviaQA | 87.6% | 73.9% |
| LMArena Expert | 1039 | — |
| BoolQ | 88.6% | — |
Multilingual Not comparable
Llama 2-70B: 27.7 (#274), phi-3-medium 14B: —
| Benchmark | Llama 2-70B | phi-3-medium 14B |
|---|---|---|
| LMArena Non-English | 1045 | — |
| LMArena Chinese | 995 | — |
| LMArena French | 1090 | — |
| LMArena German | 1041 | — |
| LMArena Japanese | 927 | — |
| LMArena Korean | 964 | — |
| LMArena Russian | 1083 | — |
| LMArena Spanish | 1143 | — |
Instruction Following Not comparable
Llama 2-70B: 54.9 (#278), phi-3-medium 14B: —
| Benchmark | Llama 2-70B | phi-3-medium 14B |
|---|---|---|
| LMArena Instruction Following | 1071 | — |
Long Context Not comparable
Llama 2-70B: 32.3 (#270), phi-3-medium 14B: —
| Benchmark | Llama 2-70B | phi-3-medium 14B |
|---|---|---|
| LMArena Longer Query | 1062 | — |
Writing & Preference Not comparable
Llama 2-70B: 32.3 (#279), phi-3-medium 14B: —
| Benchmark | Llama 2-70B | phi-3-medium 14B |
|---|---|---|
| LMArena Text | 1115 | — |
| LMArena Creative Writing | 1075 | — |
| LMArena Multi-Turn | 1088 | — |
Frequently asked questions
Is Llama 2-70B better than phi-3-medium 14B?
phi-3-medium 14B is the stronger model overall, scoring 29.7 to 24.4 on the Noometry Index.
Is Llama 2-70B or phi-3-medium 14B better for coding?
phi-3-medium 14B scores higher on coding benchmarks: 36.8 versus 31.4 in the Noometry coding category.
How many benchmarks do Llama 2-70B and phi-3-medium 14B share?
10 benchmarks have published results for both models. Llama 2-70B has 35 scored results on Noometry and phi-3-medium 14B has 13.