{"slug":"gpt-5.4","id":"gpt-5.4","type":"model","title":"GPT-5.4","description":"OpenAI's flagship model combining frontier reasoning, coding, and agentic capabilities. Unifies the best of GPT-5.3-Codex into a single model with 45% fewer hallucinations than GPT-4o.","last_updated":"2026-04-10","last_verified":null,"verification_status":"unverified","markdown_url":"/content/models/gpt-5.4.md","html_url":"/models/gpt-5.4","api_url":"/api/v1/models/gpt-5.4.json","content_hash":"dae20b16e893dcd5b0f2baeb4f785195b83ba98a1b799126bceab3845e9d7d7e","sha256":"dae20b16e893dcd5b0f2baeb4f785195b83ba98a1b799126bceab3845e9d7d7e","provider":"OpenAI","pricing":{"input":"$5.00 / 1M tokens","output":"$15.00 / 1M tokens","note":"Pricing varies by variant"},"benchmarks":{"reasoning":95,"coding":92,"math":95,"writing":93,"multilingual":90,"speed":80},"tags":["openai","proprietary","text","image","audio"],"website":"https://openai.com","release_date":"2026-03","relationships":{"links":[],"related":[{"id":"gpt-5.4-thinking","title":"GPT-5.4 Thinking","type":"model","html_url":"/models/gpt-5.4-thinking","markdown_url":"/content/models/gpt-5.4-thinking.md","shared_tags":["openai","proprietary","text","image","audio"],"score":9},{"id":"gemini-3-flash","title":"Gemini 3 Flash","type":"model","html_url":"/models/gemini-3-flash","markdown_url":"/content/models/gemini-3-flash.md","shared_tags":["proprietary","text","image","audio"],"score":6},{"id":"gemini-3.1-pro","title":"Gemini 3.1 Pro","type":"model","html_url":"/models/gemini-3.1-pro","markdown_url":"/content/models/gemini-3.1-pro.md","shared_tags":["proprietary","text","image","audio"],"score":6},{"id":"gpt-oss-120b","title":"GPT-OSS-120B","type":"model","html_url":"/models/gpt-oss-120b","markdown_url":"/content/models/gpt-oss-120b.md","shared_tags":["openai","text"],"score":6},{"id":"claude-haiku-4.5","title":"Claude Haiku 4.5","type":"model","html_url":"/models/claude-haiku-4.5","markdown_url":"/content/models/claude-haiku-4.5.md","shared_tags":["proprietary","text","image"],"score":5},{"id":"claude-opus-4.6","title":"Claude Opus 4.6","type":"model","html_url":"/models/claude-opus-4.6","markdown_url":"/content/models/claude-opus-4.6.md","shared_tags":["proprietary","text","image"],"score":5}],"explicit":{}},"metadata":{"title":"GPT-5.4","type":"model","id":"gpt-5.4","provider":"OpenAI","model_type":"proprietary","release_date":"2026-03","description":"OpenAI's flagship model combining frontier reasoning, coding, and agentic capabilities. Unifies the best of GPT-5.3-Codex into a single model with 45% fewer hallucinations than GPT-4o.","last_updated":"2026-04-10","context_window":"1M tokens","website":"https://openai.com","license":"Proprietary","modality":["text","image","audio"],"tags":["openai","proprietary","text","image","audio"],"pricing":{"input":"$5.00 / 1M tokens","output":"$15.00 / 1M tokens","note":"Pricing varies by variant"},"benchmarks":{"reasoning":95,"coding":92,"math":95,"writing":93,"multilingual":90,"speed":80},"best_for":["Complex reasoning","Coding","Multimodal analysis","Agentic workflows"]},"content_text":"# GPT-5.4\n\nOpenAI's everything model. GPT-5.4 merges the reasoning line and the coding line into a single endpoint, and the result is the most well-rounded proprietary model available. A 95/100 reasoning score and 94.6% AIME put it at or near the top of every general benchmark.\n\nThe real selling point is ecosystem. No other model has deeper integration with the tools people already use -- Microsoft 365, ChatGPT plugins, the Assistants API, and a sprawling third-party landscape. If you're building on OpenAI's platform, staying on GPT-5.4 is the path of least resistance. The 1M context window now matches Claude and Gemini, removing what used to be a disadvantage.\n\nWhere it trails: coding. At 74.9% SWE-bench, it's solid but clearly behind Claude Opus 4.6's 80.8%. The 45% hallucination reduction over GPT-4o sounds impressive until you compare it to Grok 4.20's industry-leading factual accuracy. And at $5/$15 per million tokens, it's not cheap -- Gemini 3.1 Pro delivers comparable reasoning at $2/$12.\n\n**When to pick something else:** For pure coding dominance, Claude Opus 4.6 is the better call. For budget-conscious work that doesn't sacrifice much quality, Gemini 3.1 Pro or DeepSeek V3.2 undercut GPT-5.4 significantly. If you need extended thinking for competition-level math, use GPT-5.4 Thinking instead of burning tokens on the base model.","content_length":2181,"generated_at":"2026-04-24"}