GPT-5.4

GPT-5.4 is described as a frontier model aimed at complex professional work rather than casual chat, with a single design that pulls together stronger reasoning, coding, computer use, tool workflows, spreadsheet and document handling, and long-context API use. A third-party write-up frames it as a mainline release spanning ChatGPT, the API, and Codex, with a dedicated GPT-5.4 Thinking variant for harder tasks and a gpt-5.4-pro API tier alongside the standard gpt-5.4 endpoint. The intent is to let teams hand off multi-step work involving files, code, and live tools without orchestrating every step themselves.

The model's practical strength lies in agentic coding, computer-use automation, deep research, and long-horizon tool orchestration, with positioning aimed at finance, operations, consulting, legal, and go-to-market teams that work across documents, spreadsheets, and presentations. It is described as benefiting most users who run real workflows across files and tools, while casual users may not need its depth. Native desktop and browser interaction, paired with a roughly 1M-token-class that quick-info value window for long-document and enterprise use, are presented as the main differentiators for agent builders and automation teams.

Amazon Bedrockopenai.gpt-5.4gpt

Open provider page Provider docs

Quick Info

Provider: Amazon Bedrock
Model key: openai.gpt-5.4
Release date: Mar 5, 2026
Last updated: Jun 1, 2026
Knowledge cutoff: 2025-08-31
AI SDK package: @ai-sdk/amazon-bedrock/mantle
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $2.75
Output token cost: $16.50

Limits

Output tokens: 128,000 tokens
Context window: 1,000,000 tokens

Transparent token rates

Compare GPT-5.4 pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about GPT-5.4

Abacus

CoverageBenchmark

BenchLM aggregates GPT-5.4 data as of October 17, 2026 with an aggregate capability score of 66.7/100 across 42 published benchmark rows. Its strongest category is Knowledge at rank 90th percentile, while Coding ranks 48th and Reasoning only 19th. Sustained output speed is reported at 89 tokens per second with a 1.05M-token context window. The page flags GPT-5.4 as superseded by OpenAI's newer GPT-5.5 line, framing the model as historical rather than the current flagship. Pricing is recorded at $2.50 input and $15 output per million tokens, with a $0.25 cached-input rate. Speed and benchmark evidence are linked to 21 verified and 88 public score entries drawn from mixed sources.

Abacus

CoverageBenchmark

GPT-5.4 was released on March 5, 2026 by OpenAI, according to llm-stats.com's tracked model record. The page lists OpenAI pricing at $2.50 per million input tokens and $15.00 per million output tokens, with a context window of 1.0M input and 128K output tokens. It also notes a 3.00 second p95 time-to-first-token. LLM-stats.com catalogs GPT-5.4 as a proprietary, OpenAI-released model with a 1M-token context and published leaderboard coverage across real-world prompt categories. The listing preserves OpenAI as the creator and frames the model against a broad catalog of frontier peers. Comparison entries against models like GLM-5.1 and Gemini 3 Pro appear with scores still pending.