Fast like an eagle AI at blazing speed

Latest for EAGLE

NVIDIA achieves a 1,000 TPS/user breakthrough with its new Blackwell chips using EAGLE.

NVIDIA x EAGLE

NVIDIA unveils Jetson Thor - the ultimate platform for Physical AI - delivering 3.5x faster inference using EAGLE for humanoid robotics.

NVIDIA x EAGLE

Perplexity accelerates Sonar through EAGLE for speculative decoding on their flagship LLM.

Perplexity x EAGLE

Meta achieves its fastest ever speed up for large batch sizes at production scale using EAGLE for efficient speculative decoding on Llama 4 Maverick.

Meta x EAGLE

Deepseek V3 uses EAGLE to achieve industry leading LLM speed, coined Multi-Token Prediction.

Deepseek x EAGLE
About EAGLE

EAGLE powers the AI work of leading organizations around the world

Learn More
About EAGLE

The fastest speculative decoding that exists, used by everyone

Thousands
of builders
Billions
of calls daily
Hundreds
of AI orgs

5.6x faster than vanilla decoding

The leading speed solution in autoregressive drafting with lineage across 4 EAGLE evolutions.

Explore All

Why use EAGLE

The Fastest AI

Nothing comes close to Eagle speed

The Most Coverage

Nearly all major AI models supported

No Accuracy Loss

The definition of free lunch

EAGLE is the baseline for fast decoding of LLMs, extrapolating the second-top-layer contextual feature vectors of LLMs for top speed.

EAGLE Chip - hyper AI acceleration coming soon to NVIDIA

The largest AI acceleration ever in a tiny companion chip built right into your existing GPU.

Join the waitlist

Eagle News

  • May 22, 2025

    NVIDIA Blackwell breaks the 1,000 TPS/user barrier with EAGLE

    Lorem ipsum dolor sit amet, consectetur adipiscing elit. Suspendisse varius enim in eros elementum tristique. Duis cursus, mi quis viverra ornare, eros dolor interdum nulla, ut commodo diam libero vitae erat. Aenean faucibus nibh et justo cursus id rutrum lorem imperdiet. Nunc ut sem vitae risus tristique posuere.

    Read More
  • Jun 10, 2025

    Perplexity Accelerating Sonar Through EAGLE

    Lorem ipsum dolor sit amet, consectetur adipiscing elit. Suspendisse varius enim in eros elementum tristique. Duis cursus, mi quis viverra ornare, eros dolor interdum nulla, ut commodo diam libero vitae erat. Aenean faucibus nibh et justo cursus id rutrum lorem imperdiet. Nunc ut sem vitae risus tristique posuere.

    Read More
  • Jul 1, 2025

    Fly EAGLE fly: Faster inference with vLLM & speculative decoding

    Lorem ipsum dolor sit amet, consectetur adipiscing elit. Suspendisse varius enim in eros elementum tristique. Duis cursus, mi quis viverra ornare, eros dolor interdum nulla, ut commodo diam libero vitae erat. Aenean faucibus nibh et justo cursus id rutrum lorem imperdiet. Nunc ut sem vitae risus tristique posuere.

    Read More
  • Jul 25, 2025

    LMSYS.org: Accelerating EAGLE Spec Decoding

    Lorem ipsum dolor sit amet, consectetur adipiscing elit. Suspendisse varius enim in eros elementum tristique. Duis cursus, mi quis viverra ornare, eros dolor interdum nulla, ut commodo diam libero vitae erat. Aenean faucibus nibh et justo cursus id rutrum lorem imperdiet. Nunc ut sem vitae risus tristique posuere.

    Read More
Explore All