Across 2 shared benchmarks, LlaVA-Interleave-Qwen-7B scores higher on 0 and Phi 3.5 Vision Instruct on 2. The widest gap is BLINK, where Phi 3.5 Vision Instruct scores 57 against 53.1.
| Benchmark | LlaVA-Interleave-Qwen-7B | Phi 3.5 Vision Instruct |
|---|---|---|
| BLINK | 53.1 | 57 |
| Video-MME Overall | 50.2 | 50.8 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.