Elon Musk’s SpaceXAI has launched Grok 4.7, its newest frontier artificial intelligence model, positioning it as a faster and significantly cheaper alternative to competing systems from OpenAI and Anthropic.
SpaceXAI describes Grok 4.7 as its most capable model yet for coding and knowledge work, saying it can work longer on difficult tasks, verify its own output more carefully and handle extended context more effectively than Grok 4.6. The company markets the model as delivering twice the speed at half the price of comparable AI models, although those comparisons are based on SpaceXAI’s own pricing and performance assessments rather than a universal independent benchmark.
Grok 4.7 Starts at $2 Per Million Tokens
Grok 4.7 is priced from $2 per million input tokens and $6 per million output tokens, unchanged from Grok 4.6. Cached input costs $0.50 per million tokens for prompts below 200,000 tokens, while longer prompts carry higher rates.
The model supports a 500,000-token context window and accepts both text and images as inputs. SpaceXAI is making it available through its API, Grok Build, Cursor, third-party coding systems, model routers and cloud platforms.
SpaceXAI has also introduced Grok 4.7 Fast, which runs the same underlying model using faster serving infrastructure. That version delivers roughly twice the output speed but costs twice as much and is currently limited to Cursor and Grok Build rather than the public API.
The aggressive pricing highlights intensifying competition among frontier AI developers as companies increasingly compete not only on benchmark performance but also on inference cost and speed.
SpaceXAI Targets Coding and Professional Work
Grok 4.7 uses a larger base model than Grok 4.6 and underwent a longer reinforcement-learning process focused on difficult problems requiring extended periods of work. SpaceXAI says those changes improve the model’s ability to verify its reasoning, manage longer contexts and complete complex multi-step assignments.
On CursorBench 4.0, a benchmark focused on longer-running software engineering tasks, Grok 4.7 scored 46.3%, compared with 40.4% for Grok 4.6 and 41.7% for GPT-5.6 Sol Max in SpaceXAI’s published comparison. On DeepSWE v1.1, Grok 4.7 scored 71% at high reasoning effort, versus 72.7% for GPT-5.6 Sol Max and 70% for Fable 5.1 Max.
Performance varies considerably depending on the benchmark, meaning the results do not establish one model as universally superior. SpaceXAI’s strongest relative results include electrical engineering and legal-agent tasks, while competing models remain ahead on several other evaluations.
The company is also emphasizing improvements in documents, presentations and professional knowledge work, broadening Grok’s target market beyond software developers.
AI Competition Shifts Toward Price and Performance
The launch reflects a broader shift in the AI industry. Frontier laboratories are increasingly under pressure to demonstrate that more capable models can also operate economically at enormous scale.
That matters as businesses deploy AI across coding, research, financial analysis, legal work and autonomous agents. Even relatively small differences in token prices can translate into substantial costs when companies process billions of tokens.
SpaceXAI’s strategy appears designed to compete aggressively on this dimension. Its $2 input and $6 output pricing sits substantially below several premium frontier models included in its own comparison, potentially making Grok 4.7 attractive for high-volume workloads where inference costs are critical.
At the same time, benchmark results show that the competitive landscape remains fragmented. Different frontier models lead on different categories, and real-world performance can vary depending on the task, tools and reasoning settings being used.
Musk Escalates the Frontier AI Race
Grok 4.7 arrives as SpaceXAI continues rapidly expanding both its AI models and the infrastructure supporting them. The company is competing in a market where OpenAI, Anthropic and other developers are simultaneously pushing toward more capable models designed to perform increasingly complex work with less human supervision.
For SpaceXAI, lower inference costs could become an important competitive weapon alongside raw model capability. Developers and enterprises evaluating frontier systems increasingly have to consider not simply which model performs best on a particular benchmark, but how much useful work it can complete for every dollar spent.
Grok 4.7 is SpaceXAI’s latest attempt to win that equation.
The company’s claim of twice the speed at half the cost will ultimately be tested by developers deploying the model in real workloads. But with substantially lower published token prices than several premium competitors and stronger results than Grok 4.6 across SpaceXAI’s benchmark suite, Grok 4.7 represents another escalation in the rapidly intensifying battle over the economics of frontier AI.
