Deep Cogito: Cogito v2.1 671B

● Excellent

by Deepcogito Intelligent Open Weight Rank #19 of 75

8.7
/ 10

Cogito v2.1 671B MoE represents one of the strongest open models globally, matching performance of frontier closed and open models. This model is trained using self play with reinforcement learning...

Choose a model to compare against Deep Cogito: Cogito v2.1 671B

Specifications

Specifications for Deep Cogito: Cogito v2.1 671B
AttributeValue
Lab Deepcogito
Tags Intelligent Open Weight
Overall Score 8.7/10
Release Date 2025-11
Context Window 128,000 tokens
Input Price / 1M $1.25
Output Price / 1M $1.25
Input Modalities Text
Output Modalities Text

Strengths

  • One of the strongest open models globally
  • Matches performance of frontier closed-source models
  • Self-play RL training for robust reasoning
  • 671B MoE architecture for broad capability

Weaknesses

  • Extremely large model requiring significant hardware
  • Fewer integrations than established open models
  • Limited ecosystem and community tooling

Best For

  • Cost-effective alternative to closed-source flagships
  • Self-hosted high-capability inference
  • Research and benchmarking
  • Complex reasoning and code generation

Sources & Further Reading

Related Models

Scores are aggregated from public benchmarks (MMLU, HumanEval, MATH, GSM8K, LMSYS) and normalized to a 1–10 scale. Methodology →