Research

Moonshot Kimi Model Lineage 2026

Moonshot AI Kimi release history: original K2 (Jul 2025), K2.5 multimodal (Jan 2026), K2.6 with 300-agent swarms and 256K context (Apr 2026), K3 expected June-July 2026.

By Ramanath, CTO & Co-Founder at Presenc AI · Last updated: May 2026

What this is

Moonshot AI's Kimi line went from a long-context chatbot to the best-in-class open-weight coding model in 12 months. Kimi K2.6 (April 2026) is the first open-weight model to beat GPT-5.4 (xhigh) on SWE-Bench Pro. This page is a 2026-05-15 release-by-release reference.

Kimi Release Timeline

DateReleaseNotable
Pre-2025Original Kimi chatbotLong-context (200K+) consumer assistant
Jul 2025Kimi K2 (open-source debut)1T-parameter MoE, 32B active
Jan 2026Kimi K2.5Multimodal (vision + text), 256K context
Apr 13, 2026Kimi K2.6 Code PreviewBeta release to limited testers
Apr 20, 2026Kimi K2.6 GA1T MoE, 32B active, 262K context, 300-agent swarms
Jun-Jul 2026 (expected)Kimi K3Reportedly 3-4T parameters

Kimi K2.6 Specifications

SpecValue
Total parameters1,000B (1T)
Active parameters per token32B
ArchitectureMoE: 384 experts (8 selected + 1 shared)
Layers61
Attention heads64 (Multi-head Latent Attention)
Context window262,144 tokens
Agent SwarmUp to 300 sub-agents, 4,000 coordinated steps
SWE-Bench Pro58.6 (beats GPT-5.4 xhigh at 57.7)
LicenseOpen weights

Six Things the Lineage Tells You

  1. Moonshot is the fastest-climbing open-weight lab. From product-only in late 2024 to leading open-weight coding in mid-2026.
  2. Kimi K2.6 is the first open-weight model to beat GPT-5.4 on SWE-Bench Pro. 58.6 vs 57.7 — narrow but historic.
  3. Agent Swarm at 300 sub-agents, 4,000 steps is unmatched at this scale. Most open-weight agentic frameworks cap at 10s of sub-agents.
  4. Context window grew from 200K to 262K. Smaller jumps than competitors (Llama 4 Scout: 10M, Qwen3: 256K) but consistent.
  5. Multi-head Latent Attention (MLA) is the architectural signature. Same family as DeepSeek's efficient attention.
  6. K3 expected June-July 2026 targets 3-4T parameters. Direct scale match to the frontier closed labs.

What This Means for AI Visibility

Kimi K2.6 is increasingly deployed inside agentic coding workflows because of its SWE-Bench leadership. Brands selling developer tools, libraries, or APIs should test how their products surface inside Kimi-powered coding agents. The default model behind many emerging Chinese coding assistants and on-prem enterprise deployments is now Kimi K2.6 rather than DeepSeek V4.

Methodology

Release dates and specifications from LLM Stats Kimi K2.6 page, Kimi K2.6 release blog, Shanghai NYU RITS analysis, Cloudflare Workers AI changelog, and Miraflow's K2.6 analysis.

How Presenc AI Helps

Brand-visibility tracking on Kimi-powered surfaces (Kimi chatbot, Kimi K2.6 in agentic coding tools, on-prem K2.6 deployments) is increasingly important as the model absorbs share from DeepSeek and proprietary alternatives. Presenc AI runs the same prompt suites against Kimi K2.6 as against ChatGPT, Claude, and Gemini.

Frequently Asked Questions

Moonshot AI's flagship open-weight model released April 20, 2026: 1T-parameter MoE with 32B active parameters, 262K context, 300-agent swarms with 4,000 coordinated steps, and SWE-Bench Pro 58.6 (first open-weight model to beat GPT-5.4 xhigh).
Yes from K2 onward. Original Kimi was a closed chatbot product; the K2 series (Jul 2025) marked Moonshot's open-source debut, and K2.5 / K2.6 are open weights. The next-gen K3 (expected mid-2026) is reportedly also targeting open weights but at 3-4T parameters.
Close. Kimi K2.6 leads on agentic coding (SWE-Bench Pro 58.6 vs DeepSeek V4 Pro ~83.7% on Verified is a different benchmark cut) and on multimodal. DeepSeek V4 Pro is stronger on reasoning and broader benchmarks. For agentic / multi-step coding work the current open-weight default is shifting toward Kimi K2.6.
Moonshot's coordinated multi-agent framework. K2.6 supports up to 300 sub-agents executing 4,000 coordinated steps in a single task. The capability targets long-running agentic workflows (research, coding, document processing) at a scale most open-weight frameworks can't match.

Track Your AI Visibility

See how your brand appears across ChatGPT, Claude, Perplexity, and other AI platforms. Start monitoring today.