OnlySOTA/Deep Cogito

Deep Cogito

Every model this company has on the boards we track, with each board’s scores kept apart.

1 recorded 1 with scores 0 SOTA

Models with scores

1 more (not SOTA)

Cogito v2.1

Every number here belongs to the board named beside it. OnlySOTA runs no tests and builds no composite score.

Reading How OnlySOTA reads a leaderboard A glossary for reading these boards The kill line: which AI models are worth what they cost Every company What changed

Sources Artificial Analysis LLM Leaderboard Arena — Agent Artificial Analysis Coding Agents Epoch AI Benchmarking Hub HarnessTax — SWE-bench Lite HarnessTax — Terminal-Bench 2.0 Terminal-Bench 4.0 Arena — Code Arena — Text Arena — Vision

Capabilities Coding General preference Reasoning and knowledge Agentic coding Agentic work Long context Multimodal Factuality Cost Speed