Model comparison
Deepseek Coder v2 vs phi-3-medium 14B
Deepseek Coder v2 is the stronger model overall, scoring 35.3 to 29.1 on the Noometry Index.
Last verified . 4 shared benchmarks.
DeepSeek
MicrosoftSummary
- They share 4 benchmarks with published results for both. Deepseek Coder v2 scores higher in 3 categories and phi-3-medium 14B in 0 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Deepseek Coder v2 leads 31.7 to 9.1.
- The biggest single-benchmark swing is BigCodeBench Complete: 59.7% for Deepseek Coder v2 and 48.7% for phi-3-medium 14B.
Side by side
| Deepseek Coder v2 | phi-3-medium 14B | |
|---|---|---|
| Provider | DeepSeek | Microsoft |
| Noometry Index | 35.3 | 29.1 |
| Released | 2024-06-17 | 2024-04-23 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 24 | 13 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Deepseek Coder v2 leads
Deepseek Coder v2: 37.5 (#178), phi-3-medium 14B: 36.3 (#198)
| Benchmark | Deepseek Coder v2 | phi-3-medium 14B |
|---|---|---|
| BigCodeBench Instruct | 48.2% | 37.6% |
| BigCodeBench Complete | 59.7% | 48.7% |
| LMArena Coding | 1251 | — |
| HumanEval+ | 82.3% | — |
| MBPP+ | 75.1% | — |
Reasoning Not comparable
Deepseek Coder v2: 22.6 (#175), phi-3-medium 14B: —
| Benchmark | Deepseek Coder v2 | phi-3-medium 14B |
|---|---|---|
| WinoGrande | 83.7% | 81.5% |
| LMArena Hard Prompts | 1207 | — |
| Adversarial NLI | — | 55.8% |
| BIG-Bench Hard | — | 81.4% |
| Epoch Capabilities Index | — | 121.18 |
| HellaSwag | — | 82.4% |
Math Deepseek Coder v2 leads
Deepseek Coder v2: 32.3 (#188), phi-3-medium 14B: 25.2 (#249)
| Benchmark | Deepseek Coder v2 | phi-3-medium 14B |
|---|---|---|
| LMArena Math | 1241 | — |
| MATH Level 5 | — | 17.6% |
| GSM8K | 94.5% | — |
Knowledge Deepseek Coder v2 leads
Deepseek Coder v2: 31.7 (#209), phi-3-medium 14B: 9.1 (#301)
| Benchmark | Deepseek Coder v2 | phi-3-medium 14B |
|---|---|---|
| ARC (AI2) Challenge | 64.3% | 91.6% |
| GPQA Diamond | — | 27.6% |
| LMArena Expert | 1181 | — |
| MMLU | — | 78% |
| OpenBookQA | — | 87.4% |
| TriviaQA | — | 73.9% |
Multilingual Not comparable
Deepseek Coder v2: 36.4 (#235), phi-3-medium 14B: —
| Benchmark | Deepseek Coder v2 | phi-3-medium 14B |
|---|---|---|
| LMArena Non-English | 1182 | — |
| LMArena Chinese | 1201 | — |
| LMArena French | 1184 | — |
| LMArena German | 1164 | — |
| LMArena Japanese | 1126 | — |
| LMArena Korean | 1104 | — |
| LMArena Russian | 1188 | — |
| LMArena Spanish | 1153 | — |
Instruction Following Not comparable
Deepseek Coder v2: 61.9 (#237), phi-3-medium 14B: —
| Benchmark | Deepseek Coder v2 | phi-3-medium 14B |
|---|---|---|
| LMArena Instruction Following | 1180 | — |
Long Context Not comparable
Deepseek Coder v2: 37.0 (#219), phi-3-medium 14B: —
| Benchmark | Deepseek Coder v2 | phi-3-medium 14B |
|---|---|---|
| LMArena Longer Query | 1219 | — |
Writing & Preference Not comparable
Deepseek Coder v2: 38.4 (#249), phi-3-medium 14B: —
| Benchmark | Deepseek Coder v2 | phi-3-medium 14B |
|---|---|---|
| LMArena Text | 1191 | — |
| LMArena Creative Writing | 1120 | — |
| LMArena Multi-Turn | 1176 | — |
Frequently asked questions
Is Deepseek Coder v2 better than phi-3-medium 14B?
Deepseek Coder v2 is the stronger model overall, scoring 35.3 to 29.1 on the Noometry Index.
Is Deepseek Coder v2 or phi-3-medium 14B better for coding?
Deepseek Coder v2 scores higher on coding benchmarks: 37.5 versus 36.3 in the Noometry coding category.
How many benchmarks do Deepseek Coder v2 and phi-3-medium 14B share?
4 benchmarks have published results for both models. Deepseek Coder v2 has 24 scored results on Noometry and phi-3-medium 14B has 13.