Cogito v2.1 671B is Deep Cogito's large-parameter open-weight model, with a coding index of 24.8 and an agentic index of 16.7. GPQA at 0.768 and MMLU Pro at 0.849 indicate reasonable knowledge breadth, and livecodebench at 0.688 is respectable. However, the absence of an intelligence index means overall reasoning capability cannot be fully benchmarked, and the agentic score is low for a 671B model.

For businesses, the model's knowledge scores make it potentially useful for content generation and research tasks, but the low agentic index and absence of tool use or function calling limit its applicability to automated pipelines. The 128K context window is adequate for most document tasks.

At $1.25 input and $1.25 output (symmetric pricing), cost is moderate. Without a confirmed intelligence index and with limited agentic capability, this model is best treated as a knowledge-retrieval or content-drafting tool rather than a reasoning or automation engine. Teams should test carefully before committing to production use.

Assessed June 6, 2026

Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?