GPT-5.4

Requesty published a head-to-head comparison of GPT-5.4 against xAI's grok-4.5, showing grok-4.5 winning 3 of 5 shared benchmarks. Reported scores: Intelligence Index 53.1% vs 55.8%, Coding Index 71.1% vs 72.4%, GPQA Diamond 92.0% vs 93.1%, SciCode 56.6% vs 54.1% (a GPT-5.4 win), and Humanity's Last Exam 43.7% vs 42.7% On price and context, grok-4.5 is cheaper at $2.00 input / $6.00 output per 1M tokens versus GPT-5.4's $2.50 / $15.00, but GPT-5.4 offers a larger 1.1M-token context window versus grok-4.5's 500K. Both support vision, tool calling, reasoning, and prompt caching, and both can be reached through Requesty's single OpenAI-

Requesty's AWS Bedrock listing for GPT-5.4 is pinned to the us-west-2 region and accessed as 'bedrock/gpt-5.4@us-west-2'. Bedrock's rates are slightly higher than OpenAI/Azure at $2.75 input and $16.50 output per 1M tokens, with $0.28 per 1M cache reads and a measured effective rate of $2.56 per 1M including caching. T The Bedrock page enables a clear three-way cross-cloud price comparison for engineers routing GPT-5.4: OpenAI and Azure at $2.50/$15.00 versus Bedrock at $2.75/$16.50, plus a per-provider effective rate including caching ($1.99 OpenAI, $1.54 Azure, $2.56 Bedrock). Capabilities — vision, reasoning, tool calling, caching

The Microsoft Azure AI-hosted GPT-5.4 deployment on Requesty exposes the same upstream rates as OpenAI — $2.50 input / $15.00 output per 1M tokens and $0.25 per 1M for cache reads — but Requesty measures a lower effective rate of $1.54 per 1M including caching. The endpoint id is 'azure/gpt-5.4', marked Managed EU only Requesty's production telemetry for Azure on GPT-5.4 reports a median first-token latency of 859 ms, a p95 wait of 2.81 s, and a median output throughput of 81 tokens per second, with the slowest 5% capped at 5%. Capabilities mirror the OpenAI page (vision, reasoning, tool calling, caching, web search, JSON schema), co

Requesty lists GPT-5.4 as OpenAI's latest frontier model, unifying the Codex and GPT lines into a single system. Per the vendor's pricing page (updated September 4, 2026), the upstream OpenAI deployment is accessed via model id 'openai/gpt-5.4' through the Requesty router, with input priced at $2.50 per 1M tokens, outp Capabilities shown on the OpenAI endpoint include vision, reasoning, tool calling, prompt caching, web search, and JSON schema, while computer use and image generation are marked N/A. The endpoint is labeled "Managed EU only" with "no routing and no failover," meaning it pins directly to OpenAI's deployment. Developers