OpenAI
GPT-5.4
Previous OpenAI flagship — now the value pick of the GPT-5 line at half price.
API onlytextimagefileREL 2026-03-05
Compare this model →Verified 2026-06-12
Overview
GPT-5.4 was OpenAI's flagship before 5.5 and remains excellent: ~57 on the Artificial Analysis Intelligence Index and 93.3% on GPQA Diamond, now priced at $2.50/$15 per 1M — half of GPT-5.5. Strong default for teams that want near-frontier quality at mid-tier prices.
Benchmark scores
Methodology →Intelligence57
Codingn/a
Reasoning92.8
Mathn/a
Sources and verification
- Reasoning: GPQA Diamond 92.8%
- Catalogue and pricing: OpenRouter model record · checked 2026-06-12
Strengths and weaknesses
Strengths
- Near-frontier quality at half GPT-5.5's price
- 1M+ context
- Mature ecosystem support
Trade-offs
- Superseded by 5.5 on agentic benchmarks
- Closed weights
Reach for this model when
Cost-conscious production workloads that still need top-tier reasoning.
Run something like it locally
DeepSeek V4 Pro (open weights, AA index 52) gets surprisingly close via vLLM.
