Provider ranking
Best Microsoft Models in October 2026
As of October 2026, Wizardlm 70b leads the Noometry Index with a score of 33.0, ahead of Phi 3 Medium 4k Instruct at 33.0. The two are close enough that new results could swap them.
Last verified
Every ranked Microsoft model, from the current flagship down.
- Wizardlm 70b 33.0
- Phi 3 Medium 4k Instruct 33.0
- Wizardlm 13b 31.4
- Phi 3 Mini 4k Instruct June 2024 31.3
- Phi-4 31.2
- Phi-4 Mini 30.9
- Phi 3 Mini 128k Instruct 29.7
- phi-3-medium 14B 29.7
- Phi 3 Small 8k Instruct 29.3
- Phi 3 Mini 4k Instruct 27.9
Full ranking
| 1 |  Microsoft Wizardlm 70b | Microsoft | Open | — | 33.0 | — | — | — | — |
| 2 | Microsoft Phi 3 Medium 4k Instruct | Microsoft | Open | — | 33.0 | — | — | — | — |
| 3 | Microsoft Wizardlm 13b | Microsoft | Open | — | 31.4 | — | — | — | — |
| 4 | Microsoft Phi 3 Mini 4k Instruct June 2024 | Microsoft | Open | — | 31.3 | — | — | — | — |
| 5 | Microsoft Phi-4 | Microsoft | Open | 128K | 31.2 | $0.07 | $0.14 | 357 | — |
| 6 | Microsoft Phi-4 Mini | Microsoft | Open | 128K | 30.9 | $0.075 | $0.30 | 235 | — |
| 7 | Microsoft Phi 3 Mini 128k Instruct | Microsoft | Open | — | 29.7 | — | — | — | — |
| 8 | Microsoft phi-3-medium 14B | Microsoft | Open | — | 29.7 | — | — | — | — |
| 9 | Microsoft Phi 3 Small 8k Instruct | Microsoft | Open | — | 29.3 | — | — | — | — |
| 10 | Microsoft Phi 3 Mini 4k Instruct | Microsoft | Open | — | 27.9 | — | — | — | — |
Sponsored placements are available on pages like this one. Advertise on Noometry
Compare the leaders
Frequently asked questions
Which Microsoft model is best?
As of October 2026, Wizardlm 70b leads the Noometry Index with a score of 33.0, ahead of Phi 3 Medium 4k Instruct at 33.0. The two are close enough that new results could swap them.
What are the top 5 in this ranking?
Wizardlm 70b, Phi 3 Medium 4k Instruct, Wizardlm 13b, Phi 3 Mini 4k Instruct June 2024, Phi-4, in that order, as of October 2026.
How is this list ranked?
By the Noometry Index, which is fitted from every public benchmark result and adjusted for how hard each benchmark is. Models need several independent results to be ranked.