Kimi K2.5
Moonshot AI·
LLMsopen-weightcloud + localFrontier
Overview
1T MoE (32B active) open-weight model with native multimodal vision and agent swarm support for up to 100 sub-agents.
Capabilities and innovations
1T total / 32B active MoE262K contextNatively multimodal (MoonViT 400M)Agent swarms (100 sub-agents)1,500 parallel tool callsModified MIT licenseMoonViT Vision Encoder (400M params)100 Sub-Agent Swarm Support1,500 Parallel Tool CallsTop-Ranked Artificial Analysis & LMArena at Launch
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| MMLU-Pro | 85.91% | unsourced |
| Terminal-Bench 2.x | 41.95% | independent Terminal-Bench 2.1 41.95%, measured by Vals AI with the Terminus 2 harness (accessed 2026-09-23). Vals AI is the reference harness for this field so values compare across vendors; vendor and other third-party figures stay in the notes below. Thinking mode. No Terminal-Bench 2.x value was recorded before. |
- GPQA Diamond: not reported by the vendor
- Humanity’s Last Exam (no tools): not reported by the vendor
- SWE-bench Verified: not reported by the vendor
- SWE-bench Pro: not reported by the vendor
- MMLU: not reported by the vendor
- MMMU-Pro: not reported by the vendor
Architecture and hardware
- Parameters
- 1000B total, 32B active per token (MoE)
- Estimated VRAM at Q4
- ~575 GB, Frontier class
Reliability
- Hallucination rate (Vectara HHEM)
- 14.2% (lower is better)independent
- Agentic tool use (τ-bench)
- 74.2% (higher is better)third party
Links
More from Moonshot AI
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.