SingixAI загружается…
#12LLM & ChatbotsNEWRising Star
Kimi K2.6
Moonshot AI
Триллионная MoE от Moonshot AI — Agent Swarm до 300 агентов
Launch: April 21, 2026 · Data as of June 9, 2026
88
AI Score85
PowerOverview
Kimi K2.6 (April 21, 2026) is Moonshot AI's flagship with MoE architecture: 1T total / 32B active parameters, 384 experts. 262K token context. SWE-Bench Verified 80.2%, SWE-Bench Pro 58.6% — leading open-weight coding benchmarks. Agent Swarm supports up to 300 parallel sub-agents and 4,000+ coordinated steps. 13-hour autonomous optimization demonstrated with +185% throughput improvement. Open-weight under Modified MIT. Kimi K2.7 Code (June 13, 2026) is the specialized coding variant.
Kimi K2.6 — Moonshot AI's flagship model (April 21, 2026). 1T parameter MoE (32B active, 384 experts). 262K token context. SWE-Bench Verified: 80.2%, SWE-Bench Pro: 58.6%. Agent Swarm: up to 300 parallel sub-agents, 4,000+ coordinated steps. Demonstrated 13-hour autonomous optimization (+185% throughput). Open-weight (Modified MIT). Specialized coding variant: Kimi K2.7 Code (June 13, 2026).
Capabilities
1T MoE 384 expertsAgent Swarm 300xSWE-bench 80.2%262K contextKimi K2.7 Code
Quick facts
- Best for
- Text & reasoning
- Free tier
- Free tier available
- Pricing
- Freemium
- Monthly users
- 12.0M
- Growth
- +65.0%
- Country
- China
Pricing plans
Data as of May 29, 2026. For exact prices and limits, see the official service website.
Free$0
Limited usage
ProPaid
Higher limits, priority
Capabilities
- Text / LLM
- Images
- Video
- Audio
- Code
- API
- AI agents
Integrations
Hugging FaceOllama
AI Score — breakdown
Intelligence94
Popularity76
Innovation93
Dev ecosystem89
Reliability85
Community84
Utility92
Speed87
API quality85
AI Score level: Excellent · Trend 167
Trend
Benchmarks
SWE-bench Verified: 80.2SWE-bench Pro: 58.6