EVE is now live on Wally, RunAnywhere's Jev-class decision models. Powered by Perplexity's open pplx-decider-v1-27b. Same capability, same cost, 4x faster. $5 free on signup.

We let EVE play games. Every move is one decision: it reads the screen, picks the move, and puts a probability on every option. EVE runs on @perplexity_ai's pplx-decider-v1-27b, which scores 85.71% across 11 benchmarks. Jev scores 84.51%. On Wally a decision takes 150 ms. Jev's median on JevBench is 620 ms.

Copy this into your agent to get started: "set up Then ask EVE anything with fixed answers: yes or no, pick one, or a score. You get a probability for every answer. Up to 128 questions in one request, $0.04 per million input tokens.

Model's Hugging Face: Start here:

@perplexity_ai 4x speed, zero new math. Bold bet on taste.

@perplexity_ai The per-option probabilities are the interesting bit for me. Can Wally abstain when the top choices are close, and are those scores calibrated on tasks outside the benchmark? For a product workflow, “ask a person” can be a very useful answer.

@perplexity_ai 4x faster on the same pplx-decider model is the bit that matters