Together AI· Inference platformModel update

AdapTive-LeArning Speculator System (ATLAS): A New Paradigm in LLM Inference via Runtime-Learning Accelerators

Together AI announces ATLAS, a new adaptive-learning speculator system that improves inference performance at runtime, achieving up to 500 TPS on DeepSeek-V3.1 and up to 460 TPS on Kimi-K2.

From the source

ATLAS offers a new way of doing speculative decoding — one that dynamically improves at runtime — and it fits seamlessly alongside our other Turbo techniques like the proprietary Together Turbo Speculator or Custom Speculators.

together.ai

Announced

10 Oct 2025 00:00 UTC

Detected

03 Aug 2026 21:12 UTC (+297d after)