"Moonshot AI released Kimi K2.5, their latest open-weight multimodal LLM. K2.5 excels at coding tasks, with benchmark scores comparable to frontier models such as GPT-5 and Gemini. It also features an agent swarm mode, which can direct up to 100 sub-agents for attacking problems with parallel workflow. Kimi K2.5 builds on the previous Kimi K2 MoE LLM. The new model adds vision functionality to its text-only predecessor. This in combination with its coding ability makes it excellent for front-end dev tasks."
"For the Agent Swarm feature, the Moonshot team developed a new RL technique, Parallel Agent Reinforcement Learning (PARL), to train Kimi K2.5 to decompose and parallel complex tasks. PARL was developed to address several challenges: training instability; ambiguous credit assignment; and "serial collapse", where the orchestrator simply runs a single agent. In PARL, the subagents are frozen and only the orchestrator is trained."
Moonshot AI released Kimi K2.5, an open-weight multimodal LLM with strong coding performance comparable to GPT-5 and Gemini. The model extends the Kimi K2 MoE architecture by adding MoonViT-3D vision capabilities and supports four modes: Instant, Thinking, Agent, and Agent Swarm. Agent Swarm can direct up to 100 sub-agents to decompose and execute subtasks in parallel. Kimi K2.5 continued training from a Kimi K2 checkpoint with an additional 15T tokens of pretraining, followed by supervised fine-tuning and reinforcement learning. A new PARL technique trains an orchestrator while freezing subagents to mitigate instability and credit-assignment issues.
Read at InfoQ
Unable to calculate read time
Collection
[
|
...
]