Compare / head-to-head
GPT-5.2-CodexvsGPT-5.4 mini
GPT-5.2-Codex leads 9 of 11 shared benchmarks. GPT-5.4 mini is 3x cheaper per token. Both offer a 400K-token context window.
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
9 – 2
GPT-5.2-Codex leads
Cheaper per token
GPT-5.4 mini
3x cheaper, input + output
Larger context
Tie
both 400K tokens
Providers
OpenAI
same provider
GPT-5.2-Codex
OpenAI · released 2025-12-18
textvision
GPT-5.4 mini
OpenAI · released 2026-03-17
textvision
Quality
Benchmark matrix
11 shared · 2 only GPT-5.2-Codex · 13 only GPT-5.4 mini| Benchmark | GPT-5.2-Codex | GPT-5.4 mini | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 11 | ||||
| LiveBench Agentic Coding (LiveBench) | 49.4% ↗ | 41.7% ↗ | +7.7 pt | GPT-5.2-Codex |
| LiveBench Coding (LiveBench) | 83.6% ↗ | 71.6% ↗ | +12 pt | GPT-5.2-Codex |
| LiveBench Data Analysis (LiveBench) | 78.2% ↗ | 70.8% ↗ | +7.4 pt | GPT-5.2-Codex |
| LiveBench Instruction Following (LiveBench) | 66.4% ↗ | 59.8% ↗ | +6.6 pt | GPT-5.2-Codex |
| LiveBench Language (LiveBench) | 73.7% ↗ | 71% ↗ | +2.7 pt | GPT-5.2-Codex |
| LiveBench Mathematics (LiveBench) | 88.8% ↗ | 78.5% ↗ | +10.3 pt | GPT-5.2-Codex |
| LiveBench Reasoning (LiveBench) | 77.7% ↗ | 71.3% ↗ | +6.4 pt | GPT-5.2-Codex |
| LMArena WebDev (LMArena) | 1338.9 ↗ | 1397 ↗ | -58.1 | GPT-5.4 mini |
| OSWorld-Verified | 38.2% ↗ | 72.1% ↗ | -33.9 pt | GPT-5.4 mini |
| SWE-Bench Pro | 56.4% ↗ | 54.4% ↗ | +2 pt | GPT-5.2-Codex |
| Terminal-Bench 2.0 | 64% ↗ | 60% ↗ | +4 pt | GPT-5.2-Codex |
| Only GPT-5.2-Codex reports · 2 | ||||
| Cybersecurity Capture The Flag Challenges | 67.4% ↗ | not reported | — | — |
| SWE-Lancer IC Diamond | 76% ↗ | not reported | — | — |
| Only GPT-5.4 mini reports · 13 | ||||
| GPQA Diamond | not reported | 88% ↗ | — | — |
| LMArena Elo | not reported | 1447.1 ↗ | — | — |
| Chess Puzzles (Epoch AI run) | not reported | 24% ↗epoch run | — | — |
| FrontierMath Tier 4 v2 (Epoch AI run) | not reported | 9.8% ↗epoch run | — | — |
| FrontierMath Tiers 1-3 v2 (Epoch AI run) | not reported | 51.2% ↗epoch run | — | — |
| LMArena Vision (LMArena) | not reported | 1250.8 ↗ | — | — |
| Mystery Game Puzzles (Epoch AI run) | not reported | 11% ↗epoch run | — | — |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | not reported | 88.9% ↗epoch run | — | — |
| SAGE (Vals AI) | not reported | 50.8% ↗ | — | — |
| SimpleQA Verified | not reported | 29.4% ↗epoch run | — | — |
| Terminal-Bench 4.0 (Vals AI) | not reported | 2.5% ↗ | — | — |
| Toolathlon | not reported | 42.9% ↗ | — | — |
| WeirdML (Håvard Tveit Ihle) | not reported | 60.3% ↗ | — | — |
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-5.2-Codex minus GPT-5.4 mini 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-5.2-Codex | GPT-5.4 mini | Edge |
|---|---|---|---|
| Context window | 400K tokens | 400K tokens | Tie |
| Max output | 128K tokens | 128K tokens | Tie |
| Input price / 1M | $1.75 ↗ | $0.75 ↗ | GPT-5.4 mini |
| Output price / 1M | $14 ↗ | $4.5 ↗ | GPT-5.4 mini |
| Cached input / 1M | $0.175 ↗ | $0.075 ↗ | GPT-5.4 mini |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | $15.75 | $5.25 | GPT-5.4 mini |
| Modalities | text · vision | text · vision | Tie |
| Released | 2025-12-18 | 2026-03-17 | — |
| Cited benchmark scores | 13 | 30 | — |
Reliability
OpenAI status
Live from /statusMore matchups
GPT-5.2-Codex vs …
Models sharing the most benchmarksMore matchups
GPT-5.4 mini vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run GPT-5.2-Codex and GPT-5.4 mini on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.