Grok 4.5 has delivered on Elon Musk's promise of Opus-class performance, topping Artificial Analysis's AutomationBench-AA with a 51.4% score and $0.34 per task. The independent benchmark, which spans 657 tasks across 40 simulated applications including Gmail, Slack, Salesforce, and HubSpot, confirmed that Grok 4.5 outperformed both Claude Fable 5 (48.6%) and Claude Opus 4.8 (48.5%). Grok 4.5's cost of $0.34 per task significantly undercut Fable 5's $1.35 and Opus 4.8's $1.46, with the model using approximately 8,000 output tokens per task - roughly a quarter of Opus 4.8's total. As reported by Artificial Analysis, the model's total token usage of 0.44M per task is among the lowest on the leaderboard, with low cost driven by both efficiency and competitive token pricing.
Elon Musk announced that SpaceXAI will rush Grok 4.5 to public availability tomorrow, marking an accelerated timeline from the previously confirmed launch date. According to Musk's latest announcement on social media, the model will be available to the public tomorrow following strong positive feedback from beta customers. The SpaceX and Tesla private beta program for Grok 4.5 began on June 28, 2026, with Musk stating that it performs close to or beyond Opus levels based on early evaluations. The accelerated launch timing coincides with OpenAI's planned public release of GPT-5.6 Sol, Terra, and Luna models on Thursday, creating direct competition between the two major AI companies. As reported by Rappler.com, the Trump administration had approved a broad launch of GPT-5.6 following additional testing and meetings between the company and government officials.
Grok 4.5 operates on xAI's 1.5 trillion-parameter V9 foundation, with Cursor coding data added through supplemental training. According to developer Tetsuo, the architecture is around three times larger than the v8-small model that powered Grok 4.3, making it the largest foundation model the company has released so far. The model was trained on Nvidia Blackwell GPUs at the Colossus computing cluster in Memphis. In the latest benchmark testing, Grok 4.5 completed 79.9% of task objectives and fully passed 21.9% of tasks, with particularly strong performance in Finance at 71% - ahead of Fable 5's 64% and Opus 4.8's 62%. However, the model logged 0.63 guardrail violations per task, above Opus 4.8's 0.55 and Gemini 3.5 Flash's 0.46, which could impact firms deploying agents to live financial systems where single violations incur real costs.
OpenAI announced it will expand preview access globally for GPT-5.6 models on Wednesday, according to company statements. The company had previewed its GPT-5.6 family in late June but limited access to a small group of vetted partners whose details were shared with authorities. OpenAI touted improved agentic capabilities in coding, biology and cybersecurity when it previewed the models in late June, with GPT-5.6 Sol being competitive with Anthropic's Mythos Preview on the ExploitBench cybersecurity benchmark. With both companies opening access to their latest flagship models, the focus now shifts from early previews to real-world performance as developers and enterprises begin evaluating the competing systems. The company's choice of model names attracted attention within the crypto industry because they match names of some of the crypto industry's best-known blockchain projects.
The Trump administration has established a voluntary framework under which AI developers could provide "covered frontier models" to the US government for up to 30 days before releasing them to trusted partners. This framework has significantly impacted the AI industry, with OpenAI's GPT-5.6 launch following a delay last month prompted by US government requests over heightened national security concerns about the potential misuse of powerful AI technologies. Chinese authorities are worried about the potential for Mythos to exploit software vulnerabilities and that the US might deploy the model against Beijing's interests. Anthropic had abruptly disabled its most advanced AI models, Mythos 5 and Fable 5, for all users after the US government's June 12 export control order over national security concerns, though the curbs were lifted last week after Anthropic implemented certain safeguards. The United States and China are in a race to develop cutting-edge AI models that experts have said could dramatically accelerate sophisticated cyberattacks.
