Skip to content

Compare models

Put up to 4 models beside each other — token prices, context windows, capabilities and provider, from the same catalogue the model pages read.

  1. GLM 5.3 FlashZ.AIRemove
  2. Kimi K2 0711 Preview SearchMoonshot AIRemove
  3. Muse Spark 1.3MetaRemove
glm-5.3-flash vs kimi-k2-0711-preview-search vs muse-spark-1.3
AttributeGLM 5.3 Flashglm-5.3-flashKimi K2 0711 Preview Searchkimi-k2-0711-preview-searchMuse Spark 1.3muse-spark-1.3
Pricing
Input$0.075 / 1M$0.165 / 1M$1.25 / 1M
Output$0.25 / 1M$0.49 / 1M$4.25 / 1M
Cache Write (5m)$0.075 / 1M$0.165 / 1M$1.25 / 1M
Cache Write (1h)$0.075 / 1M$0.165 / 1M$1.25 / 1M
Cache Read$0.075 / 1M$0.165 / 1M$1.25 / 1M
Web Search$0 / 1M$0 / 1M$0 / 1M
Context
Max context1M63K1M
Max outputN/AN/AN/A
Capabilities
VisionYesYesYes
Function CallingYesYesYes
JSON ModeYesYesYes
StreamingYesYesYes
Catalogue
ProviderZ.AIMoonshot AIMeta
Categorychatchatchat
Charge typePay As You GoPay As You GoPay As You Go
Released
Description
SummaryGLM-5.3-Flash is Z.AI's efficient native multimodal model, designed for coding and long-horizon agentic workflows. It combines strong multimodal capabilities with an architecture optimized for responsive, cost-efficient task execution. Built on a hybrid sparse and linear attention architecture, GLM-5.3-Flash maintains accurate long-context behavior while reducing computational overhead, making it well suited for coding agents, extended multi-step tasks, and scalable production workloads.Kimi K2 Instruct is a trillion-parameter MoE model from Moonshot AI, with 32B active parameters per step. Built for strong agentic behavior, it excels at tool use, reasoning, and code generation, leading major benchmarks in coding, logic, and tool-use tasks. It supports up to 128K context and uses a specialized training setup (including MuonClip) to stabilize very large MoE training.Muse Spark 1.3 is Meta's multimodal reasoning model designed for long-running agentic, multi-agent, and coding workflows. It maintains context and information across extended tasks, enabling reliable execution in complex, multi-step environments. The model is optimized to resolve conflicting information, seek clarification or confirmation when necessary, and execute concisely, making it well suited for autonomous agents, collaborative multi-agent systems, and long-horizon software engineering workflows.