Home / Technology / New Grok AI: Faster, Longer Tasks, No Price Hike
New Grok AI: Faster, Longer Tasks, No Price Hike
22 Sep
Summary
- Grok 4.7 offers improved coding performance without a price increase.
- The new model is designed for longer, more complex knowledge work tasks.
- Independent testing highlights token efficiency as crucial for cost.
SpaceXAI has unveiled Grok 4.7, a new AI model focused on coding and professional knowledge work. Released on 2026-09-22T00:36:16+00:00, it features enhanced benchmark performance, particularly in coding tasks like Terminal Bench. The model underwent an extended reinforcement-learning phase, resulting in improved reliability for tasks requiring several hours to complete.
Despite the enhanced capabilities, SpaceXAI has maintained the same base API pricing as the previous version, Grok 4.6. This means input and output tokens for Grok 4.7 are priced at $2/$6 USD per 1 million tokens. A faster variant is available for double the price, though its exact throughput remains unquantified.
Independent evaluations suggest that while Grok 4.7 shows gains, it still trails leading models from competitors like OpenAI and Anthropic on certain benchmarks. However, its performance on benchmarks like CursorBench has impressed some observers, particularly considering its pricing structure. The efficiency of token usage in completing tasks is emerging as a critical factor for enterprise users, potentially impacting the overall cost more than the per-token rate.
Grok 4.7 also introduces a new safeguard stack, aimed at increasing system reliability and safety. SpaceXAI reports promising results on internal safety evaluations, though these await broader independent testing. Engineering teams are advised to evaluate Grok 4.7 based on actual job completion costs and efficiency rather than solely on API list prices.
