Press Space to continue
Press Space to continue
Ask an AI model to read twice as much and it does four times the work. Hand it a 300 page book instead of one page and the attention math alone is 90,000 times heavier. That square is the reason every chatbot you’ve used gets slow and expensive the second you paste in something long. Tonight SubQ posted a latency table where the square is gone. 32K tokens in 250ms. 100K in 660ms. 384K in 2.5 seconds. I went and did the math on that last step. 100K to 384K is 3.8x the tokens and 3.8x the time,…
alex_whedon@alex_whedon3h ago
Introducing SubQ Snap, the first long-context decision model for real enterprise work. Snap is the only decision model or LLM where compute per token stays constant as context grows. • 384K tokens of multimodal context, 12x larger than Jev. • 61.7 on Decision Index v0.2.1 vs Jev’s 57.9, outperforming on 13 of 16 enterprise benchmarks. •Processes 100K tokens and returns a decision in under one second. •$0.05/MM input tokens. Output is free. Available today to anyone. The first 10,000 sign-ups get 100M free input tokens. We’ll likely hit capacity fast and will add users as we add compute. Compute offers welcome 🙂. Latency will be variable as we tune the platform to this new model and category of users. Model speeds are 80ms for minimal context, 250ms for 32k, 660ms for 100k, and 2.5 seconds for 384k. Visit: https://t.co/Nc76iJF2a8