Kimi K2.5 vs Phi-4-multimodal-instruct
Moonshot AI vs Microsoft — benchmarks, pricing, and capabilities side by side.
- •Kimi K2.5 has the higher intelligence index (87.9 vs 46.2)
- •Phi-4-multimodal-instruct is cheaper ($0.05 vs $0.40 per 1M input)
- •Kimi K2.5 is faster
- •Kimi K2.5 has a larger context window (262K)
| Kimi K2.5 | Phi-4-multimodal-instruct | |
|---|---|---|
| Intelligence index | 87.9 | 46.2 |
| Developer | Moonshot AI | Microsoft |
| Type | Multimodal | Multimodal |
| Access | Open weights | Open weights |
| Context window | 262,144 tokens | 128,000 tokens |
| Input price | $0.40 / 1M | $0.05 / 1M |
| Output price | $1.90 / 1M | $0.10 / 1M |
| Speed | 35 tok/s | 25 tok/s |
| Released | January 27, 2026 | February 1, 2025 |
| Parameters | 1T (32B active) | 5600000000 |
| Input modalities | Text, Image | — |
| Output modalities | Text | — |
Shared benchmarks
Kimi K2.5
Phi-4-multimodal-instruct
GPQA Diamond
87.9
31.5
Humanity’s Last Exam
29.4
4.4