Moonshot AI
Kimi K2 Thinking
Moonshot AI's trillion-parameter MoE reasoning model with a 256K-token context window.
Our editorial assessment as of Jun 4, 2026, based on the public sources listed below.
Overall rank
#22
Tier
A
Score
82.0
Main category
Reasoning
Strengths
- 1T/32B MoE architecture strong on BrowseComp
- Leads agentic automation and tool-use comparisons per reviewers
- Open-weight; competitive with top closed coding models on SWE-Bench Pro
Considerations
- Around 127 on Mensa-style IQ tests per trackers, below frontier vision peers
- Large hardware footprint for self-hosting
- Newer versions (K2.5/K2.6) shift the comparison frequently
Price
Open weights · API ~$0.60/$2.50 per 1M tokens
Score breakdown
Performance
84.0
Reputation
81.0
Price
80.0
Recency
88.0
Verification status
Each field carries its own re-verification date. Amber means older than 90 days — treat those figures as stale until re-checked.
Sources
- 22 AI Frontier Models Compared for 2026 - TeamAIAccessed: 2026-05-13
- The best AI models in 2026: What model to pick for your use caseAccessed: 2026-05-13