Microsoft, open weights
Phi-2
Phi-2 by Microsoft has 11 published benchmark results on Noometry, not yet enough to be ranked.
Last verified
Specifications
- Noometry rank
- Unranked
- Index score
- Not ranked
- Evidence
- 11 results
- Provider
Microsoft- Released
- December 12, 2023
- Weights
- Open weights
- Reasoning
- Unknown
- Context window
- —
- Max output
- —
- Input price
- Not listed
- Output price
- Not listed
- Blended price
- Not listed
- Output speed
- Not measured
- Value
- Not ranked
- Knowledge cutoff
- Unknown
Sponsored placements are available on pages like this one. Advertise on Noometry
Benchmark results
Every published result we track, with its source. Bold rows are the ones used for ranking; where several exist we prefer independent runs over self-reported numbers.
Coding
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| HumanEval+ | 45.1% | #35 of 45, top 78% | EvalPlus | ||
| MBPP+ | 54.2% | #31 of 38, top 82% | EvalPlus |
Reasoning
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| Adversarial NLI | 42.5% | #9 of 9, top 100% | Epoch AI | ||
| BIG-Bench Hard | 59.4% | #13 of 27, top 49% | Epoch AI | ||
| Epoch Capabilities Index | 107.94 | #193 of 213, top 91% | Epoch AI | 2023-12-12 | |
| HellaSwag | 53.6% | #28 of 29, top 97% | Epoch AI | ||
| WinoGrande | 54.7% | #42 of 43, top 98% | Epoch AI |
Knowledge
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| ARC (AI2) Challenge | 75.9% | #16 of 39, top 42% | Epoch AI | ||
| MMLU | 58.4% | #64 of 81, top 80% | Epoch AI | ||
| OpenBookQA | 73.6% | #9 of 19, top 48% | Epoch AI | ||
| TriviaQA | 45.2% | #25 of 25, top 100% | Epoch AI |
Compare Phi-2
Other Microsoft models
Frequently asked questions
How good is Phi-2?
Phi-2 by Microsoft has 11 published benchmark results on Noometry, not yet enough to be ranked.
Is Phi-2 open source?
Yes. Phi-2's weights are downloadable; check the license for commercial terms.