Cogito
Open Models with Switchable Reasoning Mode
Listed in categories:
Open SourceArtificial Intelligence
Description
Cogito v1 introduces IDA, a path to general superintelligence, with the release of the strongest LLMs in various sizes (3B, 8B, 14B, 32B, and 70B) under an open license. These models outperform existing open models in standard benchmarks, particularly the 70B model, which surpasses the Llama 4 109B MoE model. The models utilize Iterated Distillation and Amplification (IDA), a scalable alignment strategy that allows for iterative self-improvement and advanced reasoning capabilities.
How to use Cogito?
Users can download the models from platforms like Huggingface or Ollama, or access them via API on Fireworks AI or Together AI. The models can be used in standard or reasoning modes, depending on the application requirements.
Core features of Cogito:
1️⃣
Strong performance across various model sizes (3B to 70B)
2️⃣
Utilizes Iterated Distillation and Amplification (IDA) for self-improvement
3️⃣
Advanced reasoning capabilities
4️⃣
Open license for model access
5️⃣
Optimized for coding, function calling, and agentic use cases
Why could be used Cogito?
| # | Use case | Status | |
|---|---|---|---|
| # 1 | Developing AI applications that require advanced reasoning | ✅ | |
| # 2 | Creating coding assistants that can handle complex tasks | ✅ | |
| # 3 | Implementing AI in agentic roles for decision-making processes | ✅ | |
Who developed Cogito?
Deep Cogito is focused on building general superintelligence through scientific breakthroughs in advanced reasoning and iterative self-improvement. Based in San Francisco, they are backed by top venture capitalists in AI and are committed to pioneering the future of AI technology.