The AI pulse

Live headlines, new model releases and the open models developers are adopting, gathered automatically from trusted sources.

Loading the latest updates…

Latest headlines

From Google, Google DeepMind, OpenAI, Hugging Face, TechCrunch, The Verge and MIT Technology Review. Refreshed every 30 minutes.

Loading headlines…

Newest model releases

The latest models added to OpenRouter's public catalogue, which tracks hundreds of models across providers. Prices are per million tokens.

  1. Loading model releases…
Who offers the most models today

Number of models per provider in OpenRouter's live catalogue.

Model families, explained

Each major lab offers a family of models with different strengths. The descriptions are ours; the model names under each family update automatically.

Why one family has several models

Labs ship tiers because every task trades depth for speed and cost. Press play to see the difference on the same question. This is an illustration, not a benchmark.

Question: "Which parts of Garissa County are at highest flood risk this season, and what should the county do first?"
Lite tierfastest, lowest cost, best for high-volume simple tasks
SpeedDepthCost
Fast tierbalanced speed and reasoning, the everyday workhorse
SpeedDepthCost
Frontier tierdeepest reasoning for complex, high-stakes work
SpeedDepthCost

Choose a lite model for

Classifying thousands of messages, extracting fields from forms, routing support tickets, simple translations.

Choose a fast model for

Website chatbots, drafting reports and emails, summarising documents, most agent workflows.

Choose a frontier model for

Complex analysis, multi-step planning, difficult code, policy and legal drafting that people will review.

Get help choosing the right model

We benchmark models on your own data before you commit.