Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm. Built on Kimi K2 with continued pretraining over approximately 15T mixed...
Tested on 29 benchmarks · BenchGecko score 57.1. Top scores: OpenCompass — IFEval (93.9%), OTIS Mock AIME 2024-2025 (92.2%), OpenCompass — AIME2025 (91.9%).
Qwen3 Next 80B A3B Thinking scores 56.8 (99% as good) at $0.15/1M input · 67% cheaper
OpenCompass Live Code Bench v6. Fresh competitive programming problems to evaluate code generation without memorization.
Real-world software engineering tasks from GitHub issues. Models must diagnose bugs and write patches that pass test suites. Human-verified subset of SWE-bench.
Unusual and adversarial machine learning challenges. Tests robustness of reasoning about edge cases in ML systems.
Abstraction and Reasoning Corpus. Tests fluid intelligence through novel visual pattern recognition puzzles. Core measure of general intelligence.
Deceptively simple questions that humans find easy but AI models often get wrong. Tests common sense and reasoning gaps.
ARC-AGI 2, harder sequel to ARC. More complex abstract reasoning patterns that test generalization ability beyond training data.
Mock AIME (American Invitational Mathematics Exam) problems from OTIS. Tests mathematical competition performance.
OpenCompass evaluation on AIME 2025 problems. Tests mathematical reasoning on fresh competition problems.
Original research-level math problems created by professional mathematicians. Problems are unpublished and cannot be memorized.
- Typemultimodal
- Context262K tokens (~131 books)
- ReleasedJan 2026
- LicenseOpen Source
- StatusActive
- Cost / Message~$0.003
Frequently Asked Questions
Related Models
Palmyra X5 · writerDeepSeek-Coder-V2-Lite-Base · DeepSeekQwen3.8 Max (0902) · Alibaba QwenQwen3 Next 80B A3B Thinking · Alibaba QwenGrok 4.20 · xAIKey facts · as of 2026-10-05
- Kimi K2.5 by moonshotai. BenchGecko score 57.1, rank 104 of 312 scored models (normalized average of public benchmark scores).
- List price $0.45 input · $2.25 output per 1M tokens (as of 2026-10-05).
- Sold by 5 providers (as of 2026-10-05): SiliconFlow (int4) $0.45 in / $2.25 out · AtlasCloud (int4) $0.49 in / $2.50 out · Venice $0.53 in / $3.33 out · Novita $0.57 in / $2.85 out · Amazon Bedrock $0.60 in / $3.00 out. Every provider
- Gecko Tests: Who Are You B (Knows who made it) · Tokenizer Tax E (100% more tokens outside English).
How to cite · data as of 2026-10-05
Kimi K2.5 · benchmarks, pricing and providers. BenchGecko, data as of 2026-10-05. https://benchgecko.ai/model/kimi-k2-5
Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP