Compare / head-to-head
Phi-4 Multimodal InstructvsPhi-4-Reasoning-Vision-15B
Phi-4 Multimodal Instruct leads 1 of 1 shared benchmark. Phi-4 Multimodal Instruct has the larger context window (128K tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
1 – 0
Phi-4 Multimodal Instruct leads
Cheaper per token
—
list price, input + output
Larger context
Phi-4 Multimodal Instruct
128K tokens
Providers
Microsoft
same provider
Phi-4 Multimodal Instruct
Microsoft · released 2025-02-26
textvisionaudio
- Context
- 128K
- Max out
- 4K
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 19 · 0 core
Phi-4-Reasoning-Vision-15B
Microsoft · released 2026-03-04
textvision
- Context
- 16K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 10 · 0 core
Quality
Benchmark matrix
1 shared · 18 only Phi-4 Multimodal Instruct · 9 only Phi-4-Reasoning-Vision-15B| Benchmark | Phi-4 Multimodal Instruct | Phi-4-Reasoning-Vision-15B | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 1 | ||||
| OCRBench | 84.4% ↗ | 76% ↗ | +8.4 pt | Phi-4 Multimodal Instruct |
| Only Phi-4 Multimodal Instruct reports · 18 | ||||
| AI2D | 82.3% ↗ | not reported | — | — |
| BLINK | 61.3% ↗ | not reported | — | — |
| ChartQA | 81.4% ↗ | not reported | — | — |
| DocVQA | 93.2% ↗ | not reported | — | — |
| InfoVQA | 72.7% ↗ | not reported | — | — |
| InterGPS | 48.6% ↗ | not reported | — | — |
| MathVista testmini | 62.4% ↗ | not reported | — | — |
| MMBench (dev-en) | 86.7% ↗ | not reported | — | — |
| MMMU | 55.1% ↗ | not reported | — | — |
| MMMU-Pro (standard/vision) | 38.5% ↗ | not reported | — | — |
| POPE | 85.6% ↗ | not reported | — | — |
| ScienceQA Visual | 97.5% ↗ | not reported | — | — |
| Speech AI2D | 68.9% ↗ | not reported | — | — |
| Speech ChartQA | 69% ↗ | not reported | — | — |
| Speech DocVQA | 87.3% ↗ | not reported | — | — |
| Speech InfoVQA | 63.7% ↗ | not reported | — | — |
| TextVQA (val) | 75.6% ↗ | not reported | — | — |
| Video-MME (16 frames) | 55% ↗ | not reported | — | — |
| Only Phi-4-Reasoning-Vision-15B reports · 9 | ||||
| AI2D_TEST | not reported | 84.8% ↗ | — | — |
| ChartQA_TEST | not reported | 83.3% ↗ | — | — |
| HallusionBench | not reported | 64.4% ↗ | — | — |
| MathVerse_MINI | not reported | 44.9% ↗ | — | — |
| MathVision_MINI | not reported | 36.2% ↗ | — | — |
| MathVista_MINI | not reported | 75.2% ↗ | — | — |
| MMMU_VAL | not reported | 54.3% ↗ | — | — |
| MMStar | not reported | 64.5% ↗ | — | — |
| ScreenSpot_v2 | not reported | 88.2% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Phi-4 Multimodal Instruct minus Phi-4-Reasoning-Vision-15B in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Phi-4 Multimodal Instruct | Phi-4-Reasoning-Vision-15B | Edge |
|---|---|---|---|
| Context window | 128K tokens | 16K tokens | Phi-4 Multimodal Instruct |
| Max output | 4K tokens | — | — |
| Input price / 1M | — | — | — |
| Output price / 1M | — | — | — |
| Cached input / 1M | — | — | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | — | — | — |
| Modalities | text · vision · audio | text · vision | Phi-4 Multimodal Instruct |
| Released | 2025-02-26 | 2026-03-04 | — |
| Cited benchmark scores | 19 | 10 | — |
Reliability
Microsoft status
Live from /statusMore matchups
Phi-4 Multimodal Instruct vs …
Models sharing the most benchmarksMore matchups
Phi-4-Reasoning-Vision-15B vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Phi-4 Multimodal Instruct and Phi-4-Reasoning-Vision-15B on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.