
Kimi K2.6
1T open MoE agent specialist — 58.6 SWE-Bench Pro, 300-agent swarms.
About
Kimi K2.6 (April 20, 2026) is Moonshot's open-weight flagship for agentic work: a 1-trillion-parameter MoE with 32B active per token, native INT4 quantization, and a 256K context window. It scored 58.6 on SWE-Bench Pro — ahead of GPT-5.4 and Claude Opus 4.6 at release — and introduced an 'Agent Swarm' primitive coordinating up to 300 sub-agents across thousands of tool calls, with Moonshot demoing a 13-hour autonomous refactor of a financial matching engine. Weights are on Hugging Face under a Modified MIT license at $0.60/$2.50 per 1M tokens via API.
Reality check from the community: at the frontier, benchmark gaps between top models are now small enough that personal preference often matters more than spreadsheets, and K2.6 has since been superseded by Kimi K2.7 Code and Kimi K3. Self-hosting a 1T MoE is also a datacenter-scale undertaking. But for cost-sensitive, massively parallel agentic workloads, K2.6 changed the economics — roughly a fifth the token cost of the closed models it trades blows with.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 54; Trending score 90/100 from researcher assessment.