// BENCHMARK

MultiChallenge benchmark

AI model leaderboard for the MultiChallenge benchmark. Compare how large language models score on MultiChallenge, see the full ranking, and understand what this AI benchmark measures. GPT-5 currently leads with 69.6%. Multi-turn conversational benchmark testing instruction retention across 273 dialogues.

Leaderboard

# Model Organization Score Variant Source
#1 GPT-5 OpenAI 69.6% thinking-high-cited-ring-1t official ↗
#2 Grok 4 xAI 65.2% official ↗
#3 NVIDIA Nemotron 3 Ultra NVIDIA 63.8% official ↗
#4 Ling-1T Ant Group 58.24% official ↗
#5 DeepSeek-R1-0528 DeepSeek 58.2% official ↗
#6 gemini-2.5-pro Google DeepMind 57.5% official ↗
#7 Claude Sonnet 4 Anthropic 55.3% official ↗
#8 gemini-2.5-pro Google DeepMind 54.58% cited-ring-1t official ↗
#9 Kimi K2 Moonshot 54.1% official ↗
#10 MAI-Thinking-1 Microsoft 53% official ↗
#11 GLM 4.5 Zhipu AI 52.8% official ↗
#12 Qwen3-235B-A22B-Thinking (Jul 2025) Qwen 52.75% thinking-cited-ring-1t official ↗
#13 gemini-2.5-pro Google DeepMind 51.28% lowthink-cited-ling-1t official ↗
#14 Ring-1T Ant Group 50.92% thinking official ↗
#15 GPT-5 OpenAI 48.72% main-cited-ling-1t official ↗
#16 Kimi K2 Moonshot 48.72% instruct-0905-cited-ling-1t official ↗
#17 DeepSeek-V3-0324 DeepSeek 46.5% official ↗
#18 DeepSeek-V3.1-Terminus DeepSeek 45.79% thinking-cited-ring-1t official ↗
#19 MiniMax-M1-80k MiniMax 44.7% official ↗
#20 MiniMax-M1-40k MiniMax 44.7% official ↗
#21 GLM-4.5-Air Zhipu AI 42.5% official ↗
#22 DeepSeek-V3.1-Terminus DeepSeek 42.49% non-thinking-cited-ling-1t official ↗
#23 GPT-4.1 OpenAI 38.3% official ↗

Frequently asked questions about MultiChallenge

What is the MultiChallenge benchmark?

Multi-turn conversational benchmark testing instruction retention across 273 dialogues.

How is the MultiChallenge benchmark scored?

MultiChallenge is scored using the accuracy (%) metric, where a higher score is better. The maximum achievable score is 100.000. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.

Which AI model scores highest on MultiChallenge?

As of the latest reported scores on GenAIList, GPT-5 achieves the highest result on MultiChallenge with a score of 69.6%.

Is a higher MultiChallenge score better?

Yes. On MultiChallenge a higher score indicates better performance, so models near the top of the leaderboard are the strongest.