LLM ReferenceLLM Reference
All comparisons

Kimi K2.6 vs Kimi K2.5

Side-by-side comparison of specifications, capabilities, and pricing.

Released2026-04-132026-03-15
Context window262K256K
Parameters1T (MoE, 32B active)1T (MoE, 384 experts)
ArchitectureMixture of Experts (MoE)mixture of experts
LicenseProprietaryMIT
Knowledge cutoff

Capabilities

Vision
Multimodal
Reasoning
Function calling
Tool use
Structured Outputs
Code execution

Availability

Providers

Benchmarks

Google-Proof Q&A—87.9MMLU PRO—87.1Massive Multi-discipline Multimodal Understanding—84.3Instruction-Following Evaluation—93.9BFCL—68.3