Grok 4.3 (docs.x.ai)

Article summary

Grok 4.3 is a fast and reliable model with strong tool calling and instruction following capabilities, offering text and image modalities, and a context window of 1,000,000. The model is available on multiple clusters with pricing starting at $1.25 per 1M tokens. It supports function calling, structured outputs, and configurable reasoning. The model's performance and pricing are compared to other models in the industry.

Main themes

• Model Performance
• Pricing Comparison
• AI Capabilities
• Industry Benchmarking
• Model Naming
• Technology Limitations

What commenters say

• The model's speed and value are impressive, but its performance may degrade over time as seen with previous models.
• The pricing comparison with other models is not entirely accurate, and the value should be calculated based on cost per task completion.
• The model's capabilities, such as function calling and structured outputs, make it a strong contender in the industry.
• Some users are skeptical about the model's performance and question the accuracy of the benchmarks and comparisons.
• The model's naming convention is seen as unoriginal and potentially problematic, with some users suggesting that companies should pre-declare their model names to avoid conflicts.
• The technology used in the model has limitations, such as smaller context windows, which may affect its performance in real-world usage.
• Some users have tested the model and found it to be slower than expected, while others have reported satisfactory performance.
• The model's training data may include specific examples, such as pelican riding a bike, which could indicate that the model is trying to cheat or has been overfitting to certain datasets.