Model comparison
phi-3-medium 14B vs StarCoder 2 15B
phi-3-medium 14B has enough public results to be ranked (#306); StarCoder 2 15B does not yet, so treat this comparison as directional.
Last verified . 5 shared benchmarks.
Microsoft
NVIDIASummary
- They share 5 benchmarks with published results for both. phi-3-medium 14B scores higher in 1 category and StarCoder 2 15B in 0 categories, but none of those gaps is larger than the uncertainty.
- The biggest single-benchmark swing is BigCodeBench Complete: 48.7% for phi-3-medium 14B and 38.4% for StarCoder 2 15B.
Side by side
| phi-3-medium 14B | StarCoder 2 15B | |
|---|---|---|
| Provider | Microsoft | NVIDIA |
| Noometry Index | 29.7 | 38.1 |
| Released | 2024-04-23 | 2024-02-20 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 13 | 7 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
phi-3-medium 14B: 36.8 (#201), StarCoder 2 15B: 36.6
| Benchmark | phi-3-medium 14B | StarCoder 2 15B |
|---|---|---|
| BigCodeBench Complete | 48.7% | 38.4% |
| BigCodeBench Instruct | 37.6% | — |
| HumanEval+ | — | 37.8% |
Reasoning Not comparable
phi-3-medium 14B: —, StarCoder 2 15B: —
| Benchmark | phi-3-medium 14B | StarCoder 2 15B |
|---|---|---|
| Epoch Capabilities Index | 121.23 | 105 |
| WinoGrande | 81.5% | 64.3% |
| Adversarial NLI | 55.8% | — |
| BIG-Bench Hard | 81.4% | — |
| HellaSwag | 82.4% | — |
Math Not comparable
phi-3-medium 14B: 27.3 (#250), StarCoder 2 15B: —
| Benchmark | phi-3-medium 14B | StarCoder 2 15B |
|---|---|---|
| MATH Level 5 | 17.6% | — |
| GSM8K | — | 57.7% |
Knowledge Not comparable
phi-3-medium 14B: 9.1 (#306), StarCoder 2 15B: —
| Benchmark | phi-3-medium 14B | StarCoder 2 15B |
|---|---|---|
| ARC (AI2) Challenge | 91.6% | 47.2% |
| MMLU | 78% | 64.1% |
| GPQA Diamond | 27.6% | — |
| OpenBookQA | 87.4% | — |
| TriviaQA | 73.9% | — |
Frequently asked questions
Is phi-3-medium 14B better than StarCoder 2 15B?
phi-3-medium 14B has enough public results to be ranked (#306); StarCoder 2 15B does not yet, so treat this comparison as directional.
Is phi-3-medium 14B or StarCoder 2 15B better for coding?
They score almost the same on coding (36.8 vs 36.6); test both on your own repository before choosing.
How many benchmarks do phi-3-medium 14B and StarCoder 2 15B share?
5 benchmarks have published results for both models. phi-3-medium 14B has 13 scored results on Noometry and StarCoder 2 15B has 7.