Elon Musk’s X.ai has released the latest version of its AI model, Grok-1.5 vision, which has increased reasoning skills and a context length of 128,000 tokens. The new model, which will be accessible on the X platform in the days to come, promises advances in coding, math-related jobs, and long-term context awareness.
What are Grok and Grok-1.5 vision?
Grok
Grok is a generative AI chatbot created by xAI using a large language model (LLM). It was created as an effort by Elon Musk in reaction to the success of OpenAI’s ChatGPT, which Musk co-founded. The chatbot is marketed as “having a sense of humour” and providing direct access to Twitter (X). It is presently under beta testing for people using the premium version of X.
Elon Musk’s chatbot Grok has been called “anti-woke” and accused of being programmed to lie. Grok, which launched in December 2023, was discovered to provide progressive solutions on social justice, climate change, and transgender identities. However, a study discovered that its replies were marginally more left-wing and libertarian than those from ChatGPT. Musk has claimed that xAI will take urgent steps to get Grok closer to political impartiality.
According to an xAI statement, the chatbot was created to “respond to questions with a bit of wit,” have a “rebellious streak,” and be prepared to “answer spicy questions that are turned down by most other AI systems.”
Grok-1.5 vision
Grok-1.5 is the successor of Grok-1, for which X.ai only published model weights two weeks ago. The new model extends the gains achieved by its predecessor, with major increases in thinking and problem-solving abilities.
Benchmarking of Grok-1.5 vision
Grok-1.5 scored 50.6% on the MATH benchmark and 90% on the GSM8K benchmark, proving its ability to handle a wide range of competitive math problems from grade school to high school. Furthermore, the model achieved 74.1% on the HumanEval test, demonstrating its code creation and problem-solving ability.
Features of Grok-1.5 vision
One of Grok-1.5’s distinguishing features is its capacity to analyze large contexts containing up to 128K tokens within its context window. This is a huge improvement in memory capacity, permitting the model to use information from much longer papers and handle more complicated prompts while preserving its instruction-following skills. In the Needle In A Haystack (NIAH) examination, Grok-1.5 revealed powerful retrieval skills for embedded text within contexts of up to 128K tokens in length, with flawless retrieval results.
Is Gork-1.5 similar to JAX, Rust, and Kubernetes?
Built on a unique distributed training framework based on JAX, Rust, and Kubernetes, the training stack allows X.ai’s team to quickly prototype concepts and train new architectures at scale. The unique training orchestrator provides the training job’s maximum dependability and uptime by automatically detecting and ejecting faulty nodes. Checkpoints, data loading, and training task restarts have all been streamlined to reduce downtime in the event of a failure.
Conclusion
Grok-1.5 vision will shortly be accessible to early testers, and X.ai welcomes comments to help enhance the model. Several more features will be launched in the following days as the firm steadily expands the distribution of Grok-1.5.
