GPT-5.4 nano (xhigh): benchmark scores, price and ranking
GPT-5.4 nano (xhigh) from OpenAI ranks #466 of 743 LLMs on the LLMs Tiger aggregate benchmark score (47.1), based on 12 benchmark results captured up to 2026-10-13.
- Organisation
- OpenAI
- Type
- Fast / low-cost (flash) model
- Open weights
- No
- Released
- 2026-03-17 (source: epoch)
- Cheapest API price
- $0.200 in / $1.25 out per million tokens (via aihubmix, observed 2026-10-13T19:47Z)
- Aggregate rank
- #466 of 743 (score 47.1)
Relative to other models, GPT-5.4 nano (xhigh) is strongest on CritPt (#40 of 115) and weakest on LiveBench Language (#17 of 17).
Across the 12 benchmarks where at least ten models are scored, GPT-5.4 nano (xhigh) is in the top 10% on 0 and in the top half on 7.
Its cheapest listed input price, $0.200 per million tokens, is lower than that of 67% of the 427 models with a listed price.
GPT-5.4 nano (xhigh) benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| CritPt | 9.3% | #40 of 115 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LiveBench Mathematics | 91.0% | #6 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| DTBench | 80.3% | #52 of 146 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LMCA | 36.9% | #41 of 108 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LiveBench IF | 67.2% | #7 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| SciCode | 46.9% | #47 of 106 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LiveBench Agentic Coding | 46.8% | #8 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| LiveBench Reasoning | 81.1% | #9 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| ARC-AGI-2 | 5.7% | #56 of 87 | ARC Prize official leaderboard | 2026-10-13 |
| LiveBench Data Analysis | 67.6% | #14 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| LiveBench Coding | 70.8% | #15 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
| LiveBench Language | 62.5% | #17 of 17 | LiveBench 2026_06_25 official CSV | 2026-10-13 |
About GPT-5.4 nano (xhigh)
GPT-5.4 nano (xhigh) is a smaller, lower-cost model from OpenAI. It was released on 2026-03-17, according to Epoch AI. The cheapest API listing we track charges $0.200 per million input tokens and $1.25 per million output tokens, through aihubmix. We hold 12 benchmark results for it across 4 areas, and none is in the top tenth of its board.
Where GPT-5.4 nano (xhigh) stands
On our aggregate score, GPT-5.4 nano (xhigh) is #466 of 743 ranked models, in the lower half. Its score is 47.1; the scale is centred on 50, so it sits 2.9 points below the typical model. That beats 37% of the ranked field. Among the 88 ranked OpenAI models it is #62. The next one up is gpt-5.5-xhigh (codex-harness) at 47.1. The next one down is GPT-4 Turbo Preview (Nov 2023) at 46.9.
GPT-5.4 nano (xhigh), benchmark by benchmark
CritPt puts GPT-5.4 nano (xhigh) at 9.3%, #40 of 115. That beats 65% of the models tested, but 23.0 points short of GPT-5.6 Sol (max) (32.3%). The results either side are Kimi K2.7 Code (10.0%) and grok-build-0.1 (9.1%). It covers unpublished research-level physics problems. The board median is 2.9%, so GPT-5.4 nano (xhigh) is 6.4 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On LiveBench Mathematics, GPT-5.4 nano (xhigh) reaches 91.0%. It ranks #6 of 17, ahead of 65% of the field, and 4.9 points behind GPT-5.5 (xhigh) at 95.9%. The results either side are gemini-3.1-pro-preview-high (91.0%) and claude-opus-4-5-20251101-thinking-64k-high-effort (90.4%). It covers LiveBench release 2026_06_25. The board median is 88.2%, so GPT-5.4 nano (xhigh) is 2.7 points above it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
GPT-5.4 nano (xhigh) manages 80.3% on DTBench, good for #52 of 146. Claude Opus 5.5 (max) leads at 98.9%, 18.7 points further up, and 64% of models sit below. The results either side are Qwen3.5 Plus (80.5%) and Qwen3-235B-A22B-Thinking-2507 (80.3%). It covers dynamic tool-use benchmark. The board median is 68.8%, so GPT-5.4 nano (xhigh) is 11.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
LMCA puts GPT-5.4 nano (xhigh) at 36.9%, #41 of 108. That beats 62% of the models tested, but 31.4 points short of Claude Opus 5.5 (max) (68.2%). The results either side are nemotron-3-ultra (36.9%) and Qwen3.5 Plus (36.4%). It covers language-model capability assessment. The board median is 29.2%, so GPT-5.4 nano (xhigh) is 7.6 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On LiveBench IF, GPT-5.4 nano (xhigh) reaches 67.2%. It ranks #7 of 17, ahead of 59% of the field, and 11.9 points behind gemini-3.1-pro-preview-high at 79.1%. The results either side are GPT-5.4 (xhigh) (70.2%) and GPT-5.2 Codex (66.4%). It covers LiveBench release 2026_06_25. The board median is 65.2%, so GPT-5.4 nano (xhigh) is 2.0 points above it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
GPT-5.4 nano (xhigh) manages 46.9% on SciCode, good for #47 of 106. Claude Opus 5.5 (max) leads at 66.9%, 20.0 points further up, and 56% of models sit below. The results either side are MiniMax-M2.7 (47.0%) and Claude Sonnet 4.6 (max) (46.8%). It covers scientific research code. The board median is 44.3%, so GPT-5.4 nano (xhigh) is 2.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
GPT-5.4 nano (xhigh) lands below the middle on ARC-AGI-2: 5.7%, #56 of 87, with 63% of the field scoring higher. GPT-6 Astra (max) leads at 95.0%; the closest result above is o4-mini (high) at 6.1%. It covers abstract reasoning, ARC Prize public eval set. The board median is 22.8%, so GPT-5.4 nano (xhigh) is 17.1 points below it. Source: ARC Prize official leaderboard, captured 2026-10-13.
LiveBench Data Analysis is a weak spot. GPT-5.4 nano (xhigh) scores 67.6%, #14 of 17, behind 76% of the models measured. The leader, GPT-5.5 (xhigh), reaches 81.6%. It covers LiveBench release 2026_06_25. The board median is 71.8%, so GPT-5.4 nano (xhigh) is 4.1 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
LiveBench Coding is a weak spot. GPT-5.4 nano (xhigh) scores 70.8%, #15 of 17, behind 82% of the models measured. The leader, GPT-5.2 Codex, reaches 83.6%. It covers LiveBench release 2026_06_25. The board median is 76.5%, so GPT-5.4 nano (xhigh) is 5.6 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
LiveBench Language is a weak spot. GPT-5.4 nano (xhigh) scores 62.5%, #17 of 17, behind 94% of the models measured. The leader, GPT-5.5 (xhigh), reaches 87.4%. It covers LiveBench release 2026_06_25. The board median is 77.9%, so GPT-5.4 nano (xhigh) is 15.4 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
The full table above also lists the other 2 results, including boards with fewer than ten models, where a rank says little.
Strengths and weaknesses
Software engineering: uneven, ahead of 40% of the field on average across 3 benchmarks. Its best board here is SciCode (#47 of 106) and its weakest is LiveBench Coding (#15 of 17). Reasoning, maths and science: solid, ahead of 53% of the field on average across 4 benchmarks. Its best board here is CritPt (#40 of 115) and its weakest is ARC-AGI-2 (#56 of 87). General knowledge and chat: uneven, ahead of 35% of the field on average across 4 benchmarks. Its best board here is LMCA (#41 of 108) and its weakest is LiveBench Language (#17 of 17).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so GPT-5.4 nano (xhigh) costs less than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on MiniMax-M3 ($0.230) and DeepSeek-V3.1 ($0.250).
Release history
The OpenAI model released before it was GPT-5.4 mini (xhigh) on 2026-03-17, which scores 46.3. The next one was GPT-5.5 (xhigh) on 2026-04-23, at 58.0.
Reading these numbers
The 12 scores on this page come from 3 sources: LiveBench 2026_06_25 official CSV (7), Epoch AI Benchmarking Hub (4), ARC Prize official leaderboard (1). 8 of them come from an official leaderboard and 4 from datasets or third-party runs. The leaderboard also lists 5 other reasoning-effort settings of this model, which redirect here. This page shows the highest.
GPT-5.4 nano (xhigh) is also listed at 5 other reasoning-effort settings; this page shows the highest-effort row, which is the one the leaderboard keeps by default.
Models ranked nearby
- #463 Qwen1.5-110B (47.1)
- #464 gpt-5.5-xhigh (codex-harness) (47.1)
- #465 AI21-Jamba-Large-1.5 (47.1)
- #467 Kimi K2 Instruct (47.1)
- #468 Magistral Medium 1.2 (47.1)
- #469 Step 3.7 Flash (47.0)
More from OpenAI
- GPT-6 Astra (max) (#2, 64.0)
- GPT-6.1 Sol (max) (#5, 60.6)
- GPT-5.6 Sol (max) (#6, 60.1)
- GPT-6 Sol (max) (#7, 60.0)
- GPT-5.5 (xhigh) (#11, 58.0)
- GPT-5.6 Terra (max) (#12, 57.8)
Compare it on the live leaderboard → · How the aggregate score works