Skip to content

Cogito 70B

MIT

Deep Cogito · 70B · transformer-decoder

2025-03-24 131K context 70B params

Use Cases

chat code reasoning math multilingual

Quantization Options

QuantBitsVRAMQualityStatus
Q4_K_Mrec443.0 GBGood
Q8_0876.0 GBGood

About this model

Cogito 70B is the flagship model from Deep Cogito, built on a Llama 70B base with significantly enhanced reasoning capabilities. It achieves top-tier benchmark results while maintaining the MIT license for maximum flexibility. With an MMLU score of 86.0 and a 131K context window, Cogito 70B is one of the strongest open-weight reasoning models available, though it requires high-end hardware with at least 43GB of VRAM for the Q4 quantization.

Benchmarks

86.0
mmlu