New Models Available Today
OpenAI has released GPT-5.4 mini and GPT-5.4 nano, smaller variants of GPT-5.4 optimized for fast, cost-efficient inference. Both models are now available across OpenAI's API, ChatGPT, and Codex platforms.
GPT-5.4 mini delivers strong performance improvements over GPT-5 mini across multiple dimensions:
- Achieves 54.4% accuracy on SWE-Bench Pro (vs. 45.7% for GPT-5 mini)
- Runs more than 2x faster than its predecessor
- Approaches full GPT-5.4 performance on key benchmarks including OSWorld-Verified (72.1% vs. 75.0%)
- Supports the full feature set: text/image inputs, tool use, function calling, web search, file search, computer use, and skills
- 400k context window
GPT-5.4 nano serves price-sensitive use cases at just $0.20 per 1M input tokens. It's positioned for classification, data extraction, ranking, and coding subagents handling simpler supporting tasks.
Key Use Cases and Performance
The models are purpose-built for workloads where latency directly impacts product experience:
Coding Workflows: GPT-5.4 mini consistently outperforms GPT-5 mini on SWE-Bench Pro while maintaining significantly lower latency, delivering "one of the strongest performance-per-latency tradeoffs for coding workflows." It handles targeted edits, codebase navigation, front-end generation, and debugging loops effectively.
Subagents: GPT-5.4 mini excels in composite systems where larger models handle planning while smaller models execute supporting tasks in parallel (e.g., searching codebases, reviewing files, processing documents). This pattern becomes more efficient as smaller models improve.
Computer Use: On OSWorld-Verified multimodal benchmarks, GPT-5.4 mini (72.1%) substantially outperforms GPT-5 mini (42.0%) while approaching full GPT-5.4 (75.0%), making it strong for screenshot interpretation and UI interaction tasks.
Pricing and Availability
| Model | Input Cost | Output Cost | Availability |
|---|---|---|---|
| GPT-5.4 mini | $0.75 / 1M | $4.50 / 1M | API, ChatGPT, Codex |
| GPT-5.4 nano | $0.20 / 1M | $1.25 / 1M | API only |
In Codex, GPT-5.4 mini uses only 30% of the GPT-5.4 quota, allowing developers to handle simpler coding tasks for roughly one-third the cost. ChatGPT Free and Go users access mini via the Thinking feature; all other users receive it as a fallback for GPT-5.4 Thinking when rate limits are hit.
What Developers Need to Know
These models represent OpenAI's strategy of offering tiered options where "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." They're particularly valuable for applications that require responsiveness: coding assistants, real-time image analysis, and systems that parallelize work across subagents. Comprehensive benchmarks and safety information are available in OpenAI's Deployment Safety Hub.