Compare / head-to-head
MAI-Thinking-1vsPhi-4 Multimodal Instruct
MAI-Thinking-1 and Phi-4 Multimodal Instruct do not yet share a public benchmark, so quality cannot be compared apples-to-apples. MAI-Thinking-1 has the larger context window (256K tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
0
no overlap yet
Cheaper per token
—
list price, input + output
Larger context
MAI-Thinking-1
256K tokens
Provider uptime (30d)
—
Microsoft
MAI-Thinking-1
Microsoft · released 2026-08-12
text
- Context
- 256K
- Max out
- 64K
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 15 · 7 core
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
Quality
Benchmark matrix
0 shared · 15 only MAI-Thinking-1 · 19 only Phi-4 Multimodal Instruct| Benchmark | MAI-Thinking-1 | Phi-4 Multimodal Instruct | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 0 | ||||
| No public benchmark is reported for both models yet. | ||||
| Only MAI-Thinking-1 reports · 15 | ||||
| SWE-bench Verified | 73.5% ↗ | not reported | — | — |
| GPQA Diamond | 84.2% ↗ | not reported | — | — |
| MMLU-Pro | 85% ↗ | not reported | — | — |
| AIME 2025 | 97% ↗ | not reported | — | — |
| AIME 2026 | 94.5% ↗ | not reported | — | — |
| AdvancedIF | 85% ↗ | not reported | — | — |
| BFCL v3 | 72% ↗ | not reported | — | — |
| GraphWalks (<=128K) | 90% ↗ | not reported | — | — |
| HMMT February 2026 | 84.9% ↗ | not reported | — | — |
| IFBench | 69% ↗ | not reported | — | — |
| LiveCodeBench v6 | 87.7% ↗ | not reported | — | — |
| MultiChallenge | 53% ↗ | not reported | — | — |
| SimpleQA Verified | 31% ↗ | not reported | — | — |
| SWE-bench Pro | 52.8% ↗ | not reported | — | — |
| Terminal-Bench 2.0 | 46% ↗ | not reported | — | — |
| Only Phi-4 Multimodal Instruct reports · 19 | ||||
| AI2D | not reported | 82.3% ↗ | — | — |
| BLINK | not reported | 61.3% ↗ | — | — |
| ChartQA | not reported | 81.4% ↗ | — | — |
| DocVQA | not reported | 93.2% ↗ | — | — |
| InfoVQA | not reported | 72.7% ↗ | — | — |
| InterGPS | not reported | 48.6% ↗ | — | — |
| MathVista testmini | not reported | 62.4% ↗ | — | — |
| MMBench (dev-en) | not reported | 86.7% ↗ | — | — |
| MMMU | not reported | 55.1% ↗ | — | — |
| MMMU-Pro (standard/vision) | not reported | 38.5% ↗ | — | — |
| OCRBench | not reported | 84.4% ↗ | — | — |
| POPE | not reported | 85.6% ↗ | — | — |
| ScienceQA Visual | not reported | 97.5% ↗ | — | — |
| Speech AI2D | not reported | 68.9% ↗ | — | — |
| Speech ChartQA | not reported | 69% ↗ | — | — |
| Speech DocVQA | not reported | 87.3% ↗ | — | — |
| Speech InfoVQA | not reported | 63.7% ↗ | — | — |
| TextVQA (val) | not reported | 75.6% ↗ | — | — |
| Video-MME (16 frames) | not reported | 55% ↗ | — | — |
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 MAI-Thinking-1 minus Phi-4 Multimodal Instruct 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 | MAI-Thinking-1 | Phi-4 Multimodal Instruct | Edge |
|---|---|---|---|
| Context window | 256K tokens | 128K tokens | MAI-Thinking-1 |
| Max output | 64K tokens | 4K tokens | MAI-Thinking-1 |
| 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 | text · vision · audio | Phi-4 Multimodal Instruct |
| Released | 2026-08-12 | 2025-02-26 | — |
| Cited benchmark scores | 15 | 19 | — |
Reliability
Microsoft status
Last 60 days · refreshed daily on this page · live on /statusMore matchups
MAI-Thinking-1 vs …
Models sharing the most benchmarksMore matchups
Phi-4 Multimodal Instruct vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run MAI-Thinking-1 and Phi-4 Multimodal Instruct on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.