SpaceXAI Launches Grok 4.5 Targeting Software Engineering Market
PALO ALTO, Calif. — SpaceXAI launched Grok 4.5, a specialized artificial intelligence model built for complex coding, multi-step agentic workflows, and software engineering tasks.
The release targets enterprise software developers and integrates directly into development environments like Cursor and Grok Build.
SpaceXAI made the model accessible through its API platform under the identifier grok-4.5. European Union users face access delays due to regional regulatory compliance schedules.
In an official release statement, SpaceXAI detailed the scope of the new system:
"Today, we're launching Grok 4.5, SpaceXAI's smartest model built to excel at coding, agentic tasks, and knowledge work. It's our strongest model ever and was trained alongside Cursor."
SpaceXAI trained the model across tens of thousands of NVIDIA GB300 GPUs. Engineers focused on data filtering, quality scoring, and reinforcement learning tailored for multi-step engineering tasks.
According to benchmark results published by the company, Grok 4.5 recorded a 62.0 percent pass rate on DeepSWE 1.0, a benchmark that tests autonomous software problem-solving.
Fable scored 66.1 percent on DeepSWE 1.0, and OpenAI's GPT 5.5 scored 64.31 percent. Grok 4.5 scored higher than Anthropic's Claude Opus 4.8, which posted 55.75 percent.
On the SWE Marathon test, which measures multi-hour software engineering performance, Grok 4.5 reached a 29.0 percent resolution rate. Opus 4.8 recorded 26.0 percent, and Fable recorded 24.0 percent on the same test.
Describing its training focus, SpaceXAI stated:
"Our RL training covers hundreds of thousands of tasks, centered on multi-step software engineering and other technical work, with automated and model-based grading."
The company stated that Grok 4.5 can build complete web applications from single prompts using programming languages including Rust, C++, Python, and TypeScript. In demonstration tests, the model generated an interactive 3D solar system simulation using Three.js with user controls in one step.
The release follows SpaceXAI's acquisition of Cursor and the expansion of its enterprise software products. Developers can connect to Grok 4.5 through third-party coding tools or direct API calls.
Pricing for API calls positions the system alongside high-tier commercial models. Token efficiency features run during active execution loops to lower computation costs during long coding tasks.
SpaceXAI has not published specific details on dataset composition or total energy consumption for the training run.
