On Wednesday local time, Elon Musk’s technology company SpaceXAI officially released its latest large model Grok 4.5, which is the company’s first important product update since it was launched a few weeks ago. According to a blog post published by SpaceXAI that day, this new version is positioned as a "main player" for general scenarios and can undertake various tasks that the current AI industry focuses on pursuing automation, including daily knowledge-based work such as code writing and application development, office and document processing, research and writing.
SpaceXAI emphasized in its introduction that a major selling point of Grok 4.5 is “doubled token efficiency”, that is, fewer tokens are generated under the same workload, thereby significantly reducing inference costs. At a time when the price of AI model calls continues to become a core consideration for enterprise users and developers, if this efficiency advantage is verified in actual application scenarios, it will become an important bargaining chip for SpaceXAI to compete with other head models.
The company also released a set of benchmark test data to demonstrate the performance level of Grok 4.5 compared to other top models on the market. The results show that its overall capabilities have entered the first echelon. Although it is slightly lower than the current "ceiling-level" model, it is still in a highly competitive range. On the social platform
Musk posted on 4.7, but runs faster”, noting that it is the “combination of capabilities, speed and cost” that constitutes the model’s core competitiveness in the market.
In terms of price, SpaceXAI has given aggressive pricing for Grok 4.5: $2 per million input tokens and $6 per million output tokens. Combined with its claimed high efficiency, this price point is expected to become a clear attraction for cost-sensitive enterprises and developers. For comparison, Anthropic's Opus 4.7 charges about $5 per million input tokens and $25 for output tokens; OpenAI uses different model tiered pricing schemes, with the most expensive Sol version charging $5 per million input tokens and $30 for output, while the entry-level Luna model costs $1 (input) and $6 (output).
The release of Grok 4.5 coincides with this week’s “concentration of new releases” in the AI industry. Reports indicate that OpenAI plans to launch its latest and currently most powerful model, GPT 5.6, on Thursday, after the public release of the model was previously restricted due to security risk considerations by the Trump administration. OpenAI calls GPT 5.6 the company’s “strongest model to date” in its official statement. Against this background, SpaceXAI launched Grok 4.5 and tried to break through with a combination of "Opus-level capabilities + higher efficiency + lower price", making this week a new and explosive node in the large-scale model competition landscape.