@daniel_mac8 This model is not intended to be used for simple tasks (99% of tasks LLMs handle today). It has high performance when processing large amounts of context
This model has a very specific use case, donโt get caught up with the stats.
SubQ does well relative to other LLMs with large amounts of context but likely underperforms for simple atomic tasks.
Introducing SubQ - a major breakthrough in LLM intelligence.
It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA),
And the first frontier model with a 12 million token context window which is:
- 52x faster than FlashAttention at 1MM tokens
- Less than 5% the cost of Opus
Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention).
Only a small fraction actually matter.
@subquadratic finds and focuses only on the ones that do.
That's nearly 1,000x less compute and a new way for LLMs to scale.
The bond market is collapsing again.
The 10Y Note Yield is now silently back above 4.40%, the same exact level that has led to multiple market interventions by President Trump.
Simply put, the US economy cannot afford the 10Y Note Yield rising substantially above current levels.
As we saw in April 2025 and March 2026, President Trump is highly attentive to the 4.50% level on the 10Y Note Yield, which we previously labeled as our "policy pivot" point.
At the current pace, we could see 4.50%+ within a matter of days.
Meanwhile, US oil prices are above $108/barrel, gas prices are up another +5% today, and 30Y mortgage rates are at 6.50%+.
The bond market will soon become the center of attention.
Stay ahead of the trend.
3D visualization of Eulerโs formula: e^{jฮธ} = cos(ฮธ) + j sin(ฮธ).
A point traces the unit circle in the complex plane while red, green, and blue curves project the cosine, sine, and exponential paths along the ฮธ-axis.