GPT-5.4 mini
| Released | 2026-03-17 |
|---|---|
| Price per 1M | $0.75 in · $4.50 out |
| Overall score | 53.6 · rank 82 |
| Context window | 400,000 tokens |
| Provider | OpenAI |
| Weights | Proprietary |
Scores
| Dimension | Score | Rank | Benchmarks |
|---|---|---|---|
| Overall: Balanced | 53.6 | 82 | 31 |
| Overall: Coding | 53.9 | 77 | 19 |
| Overall: Reasoning | 53.1 | 91 | 15 |
| Overall: Agentic work | 53.6 | 87 | 17 |
| Overall: Frontend | 52.5 | 74 | 9 |
| Coding | 54.9 | 73 | 5 |
| Agents | 53.0 | 87 | 4 |
| Reasoning | 53.9 | 87 | 8 |
| Math | 54.6 | 75 | 4 |
| Knowledge | 48.4 | 101 | 3 |
| Frontend | 51.3 | 72 | 2 |
| General | 56.9 | 47 | 3 |
| Multimodal | 52.7 | 62 | 2 |
| Value vs price | +4.3 | 83 |
Configurations
| Configuration | Balanced | Benchmarks |
|---|---|---|
| Unspecified setting | 46.1 ±6.7 | 3 |
| High | 52.6 ±2.0 | 12 |
| Low | 44.6 ±4.8 | 5 |
| Medium | 47.7 ±8.0 | 3 |
| Non-reasoning | 44.2 ±6.0 | 5 |
| Extra High | 53.9 ±1.0 | 21 |
Evidence
Coding 54.9 ±2.0
| Benchmark | Result | Source | Date |
|---|---|---|---|
| FrontierCode | 27.0% | Epoch AI | 2026-11-11 |
| SWE-bench Verified (Vals run) | 73.0% | Vals AI | 2026-09-01 |
| SciCode | 52.1% | Epoch AI | 2026-11-11 |
| LiveCodeBench (Vals run) | 81.5% | Vals AI | 2026-09-01 |
| ProgramBench | 0.0% | Vals AI | 2026-11-09 |
Agents 53.0 ±4.4
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Terminal-Bench 4.0 (Vals run) | 2.5% | Vals AI | 2026-11-09 |
| Terminal-Bench 2.1 (Vals run) | 54.7% | Vals AI | 2026-10-01 |
| Tax Agent Bench | 8.8% | Vals AI | 2026-11-09 |
| DeepResearch Bench | 36.3% | Epoch AI | 2026-11-11 |
Reasoning 53.9 ±2.3
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Extended NYT Connections | 61.8% | Lech Mazur | 2026-11-10 |
| ARC-AGI-2 | 18.9% | Epoch AI | 2026-11-11 |
| ARC-AGI-2 (ARC Prize verified) | 18.9% | ARC Prize | 2026-11-11 |
| CritPt | 10.0% | Epoch AI | 2026-11-11 |
| ARC-AGI-1 (ARC Prize verified) | 63.7% | ARC Prize | 2026-11-11 |
| ARC-AGI-1 | 63.7% | Epoch AI | 2026-11-11 |
| GPQA Diamond | 86.9% | Epoch AI | 2026-08-07 |
| GPQA Diamond (Vals run) | 83.1% | Vals AI | 2026-09-01 |
| Kagi Reasoning Benchmark | 37.9% | Kagi | 2026-11-11 |
| Chess Puzzles | 23.0% | Epoch AI | 2026-10-13 |
| Mystery Game Puzzles | 11.0% | Epoch AI | 2026-08-27 |
Math 54.6 ±1.4
| Benchmark | Result | Source | Date |
|---|---|---|---|
| FrontierMath Tier 4 | 9.8% | Epoch AI | 2026-06-12 |
| FrontierMath Tiers 1-3 | 51.2% | Epoch AI | 2026-06-12 |
| ProofBench | 21.0% | Epoch AI | 2026-11-11 |
| OTIS Mock AIME 2024-2025 | 88.9% | Epoch AI | 2026-08-07 |
| FrontierMath Tiers 1-3 (2025 set) | 28.3% | Epoch AI | 2026-04-15 |
| FrontierMath Tier 4 (2025 set) | 2.1% | Epoch AI | 2026-04-15 |
Knowledge 48.4 ±3.6
| Benchmark | Result | Source | Date |
|---|---|---|---|
| SimpleQA Verified | 29.4% | Epoch AI | 2026-08-27 |
| MMLU-Pro (Vals run) | 84.6% | Vals AI | 2026-09-01 |
| Legal Research Bench | 12.5% | Vals AI | 2026-11-09 |
Frontend 51.3 ±0.7
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Vibe Code Bench v1.1 | 48.0% | Vals AI | 2026-11-09 |
| LMArena WebDev | 1,398 | LMArena | 2026-11-09 |
General 56.9 ±3.2
| Benchmark | Result | Source | Date |
|---|---|---|---|
| LMArena Text | 1,447 | LMArena | 2026-11-09 |
| EQ-Bench Creative Writing v3 | 1,665 | EQ-Bench | 2026-11-11 |
| EQ-Bench Longform Creative Writing | 73.0% | EQ-Bench | 2026-11-11 |
Multimodal 52.7 ±2.3
| Benchmark | Result | Source | Date |
|---|---|---|---|
| MMMU-Pro (Vals run) | 79.2% | Vals AI | 2026-09-01 |
| LMArena Vision | 1,250 | LMArena | 2026-11-09 |
Results not used in the score
| Benchmark | Result | Source | Why not used |
|---|---|---|---|
| ARC-AGI-1 | 58.0% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-1 | 13.0% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-1 | 40.8% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-1 (ARC Prize verified) | 58.0% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-1 (ARC Prize verified) | 13.0% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-1 (ARC Prize verified) | 40.8% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-2 | 13.2% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-2 | 1.1% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-2 | 4.4% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-2 (ARC Prize verified) | 13.2% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-2 (ARC Prize verified) | 1.1% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-2 (ARC Prize verified) | 4.4% | ARC Prize | A stronger configuration result was used |
| Chess Puzzles | 17.0% | Epoch AI | A stronger configuration result was used |
| Chess Puzzles | 2.0% | Epoch AI | A stronger configuration result was used |
| FrontierMath Tiers 1-3 | 17.2% | Epoch AI | A stronger configuration result was used |
| FrontierMath Tiers 1-3 | 24.6% | Epoch AI | A stronger configuration result was used |
| GPQA Diamond | 64.1% | Epoch AI | A stronger configuration result was used |
| GPQA Diamond | 83.6% | Epoch AI | A stronger configuration result was used |
| Mystery Game Puzzles | 8.0% | Epoch AI | A stronger configuration result was used |
| Mystery Game Puzzles | 7.0% | Epoch AI | A stronger configuration result was used |
| OTIS Mock AIME 2024-2025 | 26.7% | Epoch AI | A stronger configuration result was used |
| OTIS Mock AIME 2024-2025 | 87.2% | Epoch AI | A stronger configuration result was used |
Names matched to this model
gpt-5-4-mini, gpt-5-4-mini-2026-03-17