OpenAI Introduces GPT-5.4 Mini and Nano
OpenAI releases GPT-5.4 mini and nano — the most capable small models yet, bringing GPT-5.4 strengths to faster, cheaper models designed for high-volume workloads like coding assistants, subagents, and computer use.
GPT-5.4 Mini
The headline model. GPT-5.4 mini significantly improves over GPT-5 mini across coding, reasoning, multimodal understanding, and tool use — while running 2x faster. It approaches full GPT-5.4 performance on several key benchmarks:
| Benchmark | GPT-5.4 | GPT-5.4 mini | GPT-5.4 nano | GPT-5 mini |
|---|---|---|---|---|
| SWE-Bench Pro | 57.7% | 54.4% | 52.4% | 45.7% |
| Terminal-Bench 2.0 | 75.1% | 60.0% | 46.3% | 38.2% |
| Toolathlon | 54.6% | 42.9% | 35.5% | 26.9% |
| GPQA Diamond | 93.0% | 88.0% | 82.8% | 81.6% |
| OSWorld-Verified | 75.0% | 72.1% | 39.0% | 42.0% |
- Context window: 400K tokens
- Pricing: $0.75/1M input, $4.50/1M output tokens
- Available in: API, Codex, ChatGPT
GPT-5.4 Nano
The smallest, cheapest GPT-5.4 variant for speed-sensitive tasks: classification, data extraction, ranking, and coding subagents handling simpler tasks. Significant upgrade over GPT-5 nano.
Key Design Philosophy
"The best model is often not the largest one — it's the one that can respond quickly, use tools reliably, and still perform well on complex professional tasks."
This reflects a growing industry consensus: smaller, faster models are often better for production workloads where latency directly shapes user experience.
The Subagent Pattern
GPT-5.4 mini is designed for multi-model systems. In Codex, a larger model handles planning and coordination while delegating narrow subtasks (searching codebases, reviewing files, processing documents) to mini subagents running in parallel.
This "decide what to do at the top, execute quickly at scale" pattern becomes more powerful as smaller models get faster and more capable.
Computer Use
GPT-5.4 mini is particularly strong at multimodal computer use tasks — quickly interpreting screenshots of dense UIs to complete actions, approaching full GPT-5.4 performance on OSWorld-Verified.
Source: OpenAI | HN: 171 points