NewsBite analysis · AI, Startups · New Product
TechCrunch
Inherent, founded by DeepMind alumni, says its AI 'teammate' just outperformed Anthropic and OpenAI at replicating research
Inherent, a London AI lab founded by DeepMind alumni, announced that its AI agent Faraday has surpassed larger models from Anthropic and OpenAI. Using the smaller Qwen 3.6 model, Faraday managed to replicate scientific research findings independently. The approach emphasizes 'research taste,' showcasing a method of training that encourages AI agents to engage in meaningful scientific inquiry rather than simply following predefined paths.
MonitorConfidence: HighShort term horizon
Published 22 Aug 2026, 20:40 UTC4 min read
Visit original article ↗Lead image — TechCrunch
Summary
TL;DR
- Inherent's AI Faraday outperformed larger models from Anthropic and OpenAI.
- Faraday is built on the Qwen 3.6 model, significantly smaller with 27 billion parameters.
- The training method focuses on instilling 'research taste' using reinforcement learning.
- Inherent plans to grow its employee count by the end of the year.
- The company operates in-person out of its London office, fostering collaboration.
Interesting facts
- Faraday's performance is particularly notable given its smaller size compared to its competitors.
- Inherent aims to create AI capable of discovering new scientific knowledge.