Compare / head-to-head
GPT-5vs
Phi-4
GPT-5 leads 6 of 6 shared benchmarks. GPT-5 has the larger context window (400K tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
6 – 0
GPT-5 leads
Cheaper per token
—
list price, input + output
Larger context
GPT-5
400K tokens
Providers
2 providers
OpenAI · Microsoft
GPT-5
OpenAI · released 2025-08-07
textvision
Phi-4
Microsoft · released 2024-12-12
text
- Context
- 16K
- Max out
- 16K
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 17 · 7 core
Quality
Benchmark matrix
6 shared · 14 only GPT-5 · 7 only Phi-4| Benchmark | GPT-5 | Phi-4 | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 6 | ||||
| GPQA Diamond | 86.2% ↗epoch run | 56.1% ↗ | +30.1 pt | GPT-5 |
| LMArena Elo | 1434.5 ↗ | 1216.6 ↗ | +217.9 | GPT-5 |
| BALROG (BALROG) | 32.8% ↗ | 11.6% ↗ | +21.2 pt | GPT-5 |
| Chess Puzzles (Epoch AI run) | 37% ↗epoch run | 1% ↗epoch run | +36 pt | GPT-5 |
| MATH Level 5 (Epoch AI run) | 98.1% ↗epoch run | 64.9% ↗epoch run | +33.2 pt | GPT-5 |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | 91.4% ↗epoch run | 13.8% ↗epoch run | +77.6 pt | GPT-5 |
| Only GPT-5 reports · 14 | ||||
| SWE-bench Verified | 74.9% ↗ | not reported | — | — |
| AIME 2025 | 94.6% ↗ | not reported | — | — |
| EBR-bench (Epoch AI run) | 12.7% ↗epoch run | not reported | — | — |
| FrontierMath Tier 4 v2 (Epoch AI run) | 22% ↗epoch run | not reported | — | — |
| FrontierMath Tiers 1-3 v2 (Epoch AI run) | 55.4% ↗epoch run | not reported | — | — |
| GSO Opt@1 (GSO) | 5.9% ↗ | not reported | — | — |
| LMArena Vision (LMArena) | 1210.2 ↗ | not reported | — | — |
| LMArena WebDev (LMArena) | 1417.1 ↗ | not reported | — | — |
| Mystery Game Puzzles (Epoch AI run) | 23% ↗epoch run | not reported | — | — |
| SAGE (Vals AI) | 43.7% ↗ | not reported | — | — |
| SimpleBench (SimpleBench) | 56.7% ↗ | not reported | — | — |
| SimpleQA Verified | 50.1% ↗epoch run | not reported | — | — |
| SWE-bench Verified (Epoch AI run) | 73.6% ↗epoch run | not reported | — | — |
| WeirdML (Håvard Tveit Ihle) | 60.7% ↗ | not reported | — | — |
| Only Phi-4 reports · 7 | ||||
| MMLU | not reported | 84.8% ↗ | — | — |
| MMLU-Pro | not reported | 70.4% ↗ | — | — |
| HumanEval | not reported | 82.6% ↗ | — | — |
| ArenaHard | not reported | 75.4% ↗ | — | — |
| HumanEval+ | not reported | 82.8% ↗ | — | — |
| IFEval | not reported | 63% ↗ | — | — |
| MATH | not reported | 80.4% ↗ | — | — |
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 GPT-5 minus Phi-4 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 | GPT-5 | Phi-4 | Edge |
|---|---|---|---|
| Context window | 400K tokens | 16K tokens | GPT-5 |
| Max output | 128K tokens | 16K tokens | GPT-5 |
| Input price / 1M | $1.25 ↗ | — | — |
| Output price / 1M | $10 ↗ | — | — |
| Cached input / 1M | $0.125 ↗ | — | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | $11.25 | — | — |
| Modalities | text · vision | text | GPT-5 |
| Released | 2025-08-07 | 2024-12-12 | — |
| Cited benchmark scores | 27 | 17 | — |
Reliability
Provider status
Live from /statusMore matchups
GPT-5 vs …
Models sharing the most benchmarksMore matchups
Phi-4 vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run GPT-5 and Phi-4 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.