SpaceXAI Unveils Grok 4.5: A Faster, High-Efficiency 'Opus-Class' Competitor
小葵API服务 的 AI API 使用建议
小葵API服务 面向需要 OpenAI 兼容接口、Claude/Gemini/GPT 多模型切换、包月额度管理和图像模型调用的用户。阅读本文后，可以结合本站的模型清单、独立使用文档和个人面板，把教程内容直接落到实际调用流程中。
The landscape of generative artificial intelligence is shifting once again. In its first major product release since going public, Elon Musk's SpaceXAI has officially launched Grok 4.5.
Positioned as a highly efficient "workhorse" model, Grok 4.5 is designed to tackle intensive, routine knowledge work—including coding, application development, research, writing, and clerical tasks—at a fraction of the cost of its nearest competitors.
Under the Hood: The 1.5T V9 Architecture
Initially teased in late June during a private beta, Grok 4.5 is built on SpaceXAI's massive 1.5-trillion parameter V9 model architecture. This update integrates advanced features like Cursor training and a specialized reinforcement learning framework known as the Grok Build RL harness.
According to SpaceXAI, the primary differentiator for Grok 4.5 isn't just raw intelligence, but its "twice greater token efficiency" compared to other leading systems. In the AI industry, where the recurring costs of API tokens can quickly spiral for enterprise consumers, doubling efficiency represents a major competitive advantage.
An "Opus-Class" Model at a Fraction of the Cost
Elon Musk took to his social media platform X (a subsidiary of SpaceXAI) to champion the capabilities of the new release, directly comparing it to Anthropic’s flagship model, Claude 3 Opus:
"Based on strong positive feedback from customers in our beta test program, @SpaceXAI will make Grok 4.5 available to the public tomorrow. It is an Opus-class model, but faster, more token-efficient and lower cost."
Musk later added that SpaceXAI's internal benchmarks put Grok 4.5 roughly on par with the capabilities of Anthropic's newer Opus 4.7, while operating at much higher speeds.
How the Pricing Compares
The pricing structure released by SpaceXAI positions Grok 4.5 as a highly disruptive force in the enterprise AI market:
| Model | Input Cost (per 1M tokens) | Output Cost (per 1M tokens) |
|---|---|---|
| Grok 4.5 | $2.00 | $6.00 |
| Claude Opus 4.7 | $5.00 | $25.00 |
| OpenAI Sol (Premium) | $5.00 | $30.00 |
| OpenAI Luna (Lightweight) | $1.00 | $6.00 |
At just $2 per million input tokens and $6 per million output tokens, Grok 4.5 offers near-premium intelligence at prices closer to competitors' budget tiers.
Benchmark Performance and Market Impact
SpaceXAI released a series of benchmark metrics showing that Grok 4.5 is highly competitive with other top-tier models, landing just shy of absolute best-in-class performance but dominating on speed and cost metrics.
This release kicks off an incredibly busy week for the AI industry. As SpaceXAI rolls out Grok 4.5, OpenAI is preparing to counter with the launch of GPT 5.6, a model previously delayed due to security reviews by the Trump administration.
With major players rapidly lowering API pricing while squeezing more performance out of their architectures, developers and enterprises are the ultimate winners in this ongoing AI price war. It remains to be seen how developers will utilize Grok's new RL-harness capabilities in real-world deployments, but the initial outlook for SpaceXAI's first public-facing model is highly promising.