OpenAI released GPT-5.4 yesterday. Today, you can use it in Frontman.
GPT-5.4 is OpenAI’s most capable model for professional work. It combines the coding strengths of GPT-5.3-Codex with improved reasoning, native computer-use capabilities, and a 1M token context window. For Frontman users, this means longer sessions, more complex tasks, and fewer interruptions to manage context.
What GPT-5.4 Brings to Frontman
1M token context window. GPT-5.4 supports up to one million tokens of context. That is roughly 750,000 words — enough to hold your entire codebase, conversation history, and documentation in a single session. No more trimming context or losing track of earlier decisions.
Native computer-use capabilities. GPT-5.4 is the first general-purpose model with built-in computer-use abilities. It can operate browsers, navigate desktop environments, and execute multi-step workflows across applications. On OSWorld-Verified, it achieves 75.0% — surpassing human performance at 72.4%.
Tool search for large ecosystems. When working with many MCP servers, GPT-5.4 uses tool search to find and invoke the right tools without loading every definition upfront. In OpenAI’s testing, this reduced token usage by 47% while maintaining the same accuracy. Faster responses, lower costs.
Token-efficient reasoning. GPT-5.4 uses significantly fewer tokens to solve problems compared to GPT-5.2. The same intelligence, fewer tokens, faster speeds. For long-running agentic tasks, this adds up.
Frontier coding. GPT-5.4 incorporates the coding capabilities of GPT-5.3-Codex while matching or exceeding it on SWE-Bench Pro. It excels at complex frontend tasks with more aesthetic and functional results than previous models.
How to Use It
GPT-5.4 is now the default model for ChatGPT OAuth users in Frontman. If you have connected your ChatGPT Pro or Plus account, you will see GPT-5.4 in the model picker immediately.
For OpenRouter users, both GPT-5.4 and GPT-5.4 Pro are available — select either from the model dropdown. GPT-5.4 Pro scores 89.3% on BrowseComp (web research), 83.3% on ARC-AGI-2 (abstract reasoning), and 58.7% on Humanity’s Last Exam with tools. If you need maximum performance on the hardest problems, Pro is the option.
No configuration changes required. Open Frontman, pick the model, and start working.
The Bigger Picture
Every major model release makes AI coding agents more capable. GPT-5.4 is notable because it closes gaps that mattered: longer context means fewer session restarts, native computer-use means more complete automation, tool search means practical scalability for real MCP setups.
Frontman connects these capabilities to your actual browser and source code. The model sees what you see. It edits the files you are working on. When GPT-5.4 reasons through a complex task, Frontman makes sure that reasoning lands in the right place.
Try it at frontman.sh.