Kimi K2 Thinking: clarified the AIME benchmark label as AIME 2025

The OnAirToday AI Progress Tracker recorded a verified change to Moonshot's Kimi K2 Thinking (version kimi-k2-thinking): Clarified the AIME benchmark label as AIME 2025.

Kimi K2 Thinking is Moonshot AI's open-weight reasoning model for text-based analysis, coding and multi-step tool use, with a documented 256K-token context window. Its published benchmark results depend on evaluation settings and tool access; results and security claims concerning other Kimi versions do not establish equivalent performance or vulnerabilities in this model.

What changed in this review:

• Removed Terminal-Bench and FrontierMath figures without established published results attributable specifically to Kimi K2 Thinking.
• Kimi K2.5 is a newer family release; its capabilities and benchmark results are not attributed to Kimi K2 Thinking.
• Added family-level reporting of a Kimi K3 Redis exploit; it does not establish a Kimi K2 Thinking vulnerability.
• Capability score 65 → 78
• Flaws and limitations updated
• Risks and threats updated

Teams building on Kimi K2 Thinking should confirm limits, pricing and benchmark coverage against the vendor's own documentation before relying on these figures.

The full profile — capabilities, benchmark citations, known flaws and risk guidance — is on the tracker at /ai-models/kimi-k2-thinking.

References:

• https://huggingface.co/moonshotai/Kimi-K2-Thinking
• https://github.com/MoonshotAI/Kimi-K2
• https://platform.moonshot.ai/docs/pricing/chat