Inception Launches 'Mercury 3' with Speeds Up to 8 Times Faster Than Competitors
Inception has launched the 'Mercury 3' diffusion model with a context window of one million tokens, claiming speeds up to 8 times faster than competitors without publishing independent results.
Inception has launched the "Mercury 3" model, a language model based on diffusion technology with a context window of up to one million tokens, according to a report by SiliconANGLE.
The company stated that the model recorded 1,068 tokens per second on Nvidia chips, compared to about 240 for "Claude Haiku 5.5" and 130 for "GPT-6 Luna" in its internal tests.
It added that it outperformed in 8 out of 12 tests it conducted, but trailed in the "DeepSWE" benchmark, according to the figures it released.
The company has not published independent results verifying these figures, meaning the comparison currently relies solely on its own proprietary data.
What Do These Terms Mean?
Diffusion Model: A type of artificial intelligence model that generates text or images through gradual refinement rather than generating word by word.
Context: The maximum amount of tokens that the model can read and process in a single request.
Token: The smallest unit of text processed by the model, which can be a word, part of a word, or a punctuation mark.
Benchmark: A standardized set of tasks used to measure a model's capabilities and compare its performance against other models.
Weekly Newsletter
Read between the lines before everyone else. Decode the most important economic, tech, and decision-maker movements in the region.. in 5 minutes every Saturday.
