@marktenenholtz
Member of Technical Staff @perplexity_ai
AI
Perplexity says the model and agent run on-device, while its separate Hybrid mode can use cloud models.
AI
The user says the evaluation was for a personal project and that decision models have a long way to go.
AI
A quote post says the task is to forecast crypto coins with a holdout set from 2024.
AI
Artificial Analysis says Exa's auto mode with Dynamic Highlights scored 80 on its Search Index, tying Perplexity Search (medium).
AI
One post argues raw-data brokerage will commoditize, while refined post-training tasks will grow more differentiated.
AI
OpenAI Devs says the dot investigated the issue and started a Codex thread he could steer from his phone.
AI
Perplexity says Pro and Max subscribers have access; a separate post says the model is available to all users.
AI
Perplexity's README says SPACE uses isolated Firecracker microVMs and an HTTP API to run commands and collect output.
AI
A user says a student from Waterloo sent a templated AI-generated response after the user replied to an advice request.
AI
A user says they're playing around with the model and links to a photo.
AI
A user praises what people are building with EmbeddingGemma 2 and welcomes the attention embedding models are getting.
AI
A post argues that human tasks are now too easy to use as AI benchmarks in many domains.
AI
Reuters, citing the Financial Times, says Nvidia may buy Reflection AI or invest further after an earlier $800 million stake.
AI
A user jokes that models using linear attention once had names like Linformer, but now they’re called Qwen3.8.
AI
A user says many investors’ surprise shows how little they understand about models, training and data.
AI
Perplexity says embedding images and rendered pages directly preserves tables, figures and layout that text extraction drops.
AI
Perplexity says its new pplx-decider-v1.1-27b model costs $0.02 per million input tokens, half the price of v1. OpenRouter later said the model was also available there with text, JSON, or image inputs and free output.
AI
Reflection says Beam uses 23B active parameters and plans to publish its full weights later in October 2026.
AI
An employee marked their first year at Perplexity by praising the team.
AI
A commenter says the paper leaves “single-shot accuracy” unclear and omits key evaluation settings.
