GPT-5.4 Mini & Nano Launch: Lightweight, Cost-Effective Models Built for Scale

​

Key Highlights

• Most Capable Small Models: GPT-5.4 Mini significantly improves over GPT-5 Mini across coding, reasoning, multimodal understanding, and tool use — over 2x faster
• Ultra Cost-Effective: GPT-5.4 Nano at just $0.20/M input tokens and $1.25/M output tokens, the cheapest GPT-5.4 family model
• Near-Flagship Performance: Mini scores 54.4% on SWE-Bench Pro and 72.1% on OSWorld-Verified, approaching full GPT-5.4
• 400K Context Window: Consistent with the GPT-5 family
• Full Capabilities: Text & image input, tool calling, web search, computer use — all supported

​

Background

​

Detailed Analysis

​

Core Features

​

GPT-5.4 Mini Benchmarks

​

GPT-5.4 Nano Performance

​

Adjustable Reasoning Effort

• none: No reasoning, fastest response
• low / medium / high: Increasing reasoning depth
• xhigh: Maximum reasoning effort

​

Practical Applications

​

Recommended Use Cases

1. Coding Assistant: 54.4% SWE-Bench Pro, ideal for code generation, review, and debugging
2. Autonomous Agents: 72.1% OSWorld, supports computer use and multi-step workflows
3. Daily Chat: Default thinking model for ChatGPT Free and Go users
4. OpenClaw & Similar: High performance at low cost, perfect for batch intelligent tasks

1. Data Pipelines: Classification, extraction, ranking at high throughput
2. Coding Subagents: Handle simpler supporting tasks in agent architectures
3. Batch Image Understanding: ~$52 for 76,000 photo descriptions
4. Real-time Classification: Ultra-low latency, ultra-low cost

​

Code Examples

​

Using GPT-5.4 Mini

from openai import OpenAI

client = OpenAI(
    api_key="your-apiyi-key",
    base_url="https://api.apiyi.com/v1"
)

# GPT-5.4 Mini - Cost-effective coding assistant
response = client.chat.completions.create(
    model="gpt-5.4-mini",
    messages=[
        {
            "role": "user",
            "content": "Refactor this code to extract common logic and add error handling..."
        }
    ],
    max_tokens=4096
)

print(response.choices[0].message.content)

​

Using GPT-5.4 Nano for Batch Processing

# GPT-5.4 Nano - Ultra cost-effective batch tasks
response = client.chat.completions.create(
    model="gpt-5.4-nano",
    messages=[
        {
            "role": "system",
            "content": "You are a text classifier. Classify user input as: positive, negative, or neutral. Output only the classification."
        },
        {
            "role": "user",
            "content": "This product is excellent quality, will definitely buy again!"
        }
    ],
    max_tokens=10
)

print(response.choices[0].message.content)

​

Pricing & Availability

​

Pricing

​

Promotions

​

Available Models

​

How to Access

• Website: apiyi.com
• API Endpoint: https://api.apiyi.com/v1
• OpenAI-compatible format, works with all OpenAI SDKs
• Official API proxy, stable and reliable

​

Summary & Recommendations

• Mini: Flagship-tier capability at 30% cost, 2x speed
• Nano: Ultra-low pricing at 8% of flagship cost, built for scale

1. Near-flagship performance needed: Choose GPT-5.4 Mini (SWE-Bench Pro 54.4%, OSWorld 72.1%)
2. High-throughput batch processing: Choose GPT-5.4 Nano ($0.20/M input tokens)
3. Agent architecture subtasks: Nano as execution layer, Mini as decision layer
4. OpenClaw & similar: Both Mini and Nano work well — choose based on cost-performance needs
5. Cost-insensitive professional tasks: Still recommend GPT-5.4 flagship or Pro