5 text
DeepSeek's cost-efficient V4.1 tier and the first model built on the Causal Encoder-Decoder (CED) architecture: a 552B MoE activating 8B parameters on input and 16B on output.
Z.ai's large-scale reasoning flagship for complex software engineering and long-horizon agent tasks.
The GA release of DeepSeek's efficiency-optimized V4 Flash, re-post-trained for coding, reasoning, and agent workflows.
MoonshotAI's 2.8T-parameter flagship for complex coding, knowledge work, and long-horizon agent runs.
Efficient Mixture-of-Experts Gemma 4 with only 4B active parameters per token.