Model card
GPT-5.4.
OpenAIapi
Imported from https://raw.githubusercontent.com/GAIR-NLP/AcademiClaw/main/README.md
§ 02 · Benchmarks
Every benchmark GPT-5.4 has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | AcademiClaw | Agentic AI · Task agents | safety-score | 87.5% | #2 | 2026-05-04 | source ↗ |
| 02 | React Native Evals | Mobile Development · React Native Code Generation | navigation-satisfaction | 95.6% | #2 | — | source ↗ |
| 03 | React Native Evals | Mobile Development · React Native Code Generation | async-state-satisfaction | 85.4% | #2 | — | source ↗ |
| 04 | AcademiClaw | Agentic AI · Task agents | avg-score | 65.6% | #3 | 2026-05-04 | source ↗ |
| 05 | React Native Evals | Mobile Development · React Native Code Generation | requirement-satisfaction | 82.6% | #3 | — | source ↗ |
| 06 | React Native Evals | Mobile Development · React Native Code Generation | animation-satisfaction | 68.9% | #3 | — | source ↗ |
| 07 | AcademiClaw | Agentic AI · Task agents | pass | 42.5% | #4 | 2026-05-04 | source ↗ |
| 08 | AcademiClaw | Agentic AI · Task agents | avg-time-sec | 240.00 | #5 | 2026-05-04 | source ↗ |
| 09 | AcademiClaw | Agentic AI · Task agents | avg-tokens-per-task-k | 525.00 | #6 | 2026-05-04 | source ↗ |
| 10 | AcademiClaw | Agentic AI · Task agents | tool-calls-per-task | 19.0% | #6 | 2026-05-04 | source ↗ |
| 11 | HLE | Reasoning · Multi-step Reasoning | accuracy | 36.2% | #9 | — | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 03 · Strengths by area
Where GPT-5.4 actually performs.
§ 04 · Papers
1 paper with results for GPT-5.4.
- 2026-05-04· Agentic AI· 6 results
AcademiClaw: When Students Set Challenges for AI Agents
Junjie Yu, Pengrui Lu, Weiye Si, Hongliang Lu et al.
§ 05 · Related models
Other OpenAI models scored on Codesota.
§ 06 · Sources & freshness
Where these numbers come from.
paper
6
results
Callstack Incubator
4
results
scale-hle-official
1
result
11 of 11 rows marked verified.