BABILong (NIAH, v0, 4k) Leaderboard · The Aggregate

BABILong (NIAH, v0, 4k): leaderboard

Metric: QA1–5 Mean Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 29 models tracked.

Top models

#	Model	Score
1	Llama 3.1 70B Instruct	74
2	GPT-4 Preview (0125)	74
3	Llama 3.1 8B Instruct	66
4	Mixtral 8x22B Instruct (v0.1)	65
5	Phi-3-medium-128k-instruct	62
6	c4ai-command-r-v01	61
7	Mixtral 8x7B Instruct (v0.1)	55
8	Yi-34B-200K	54
9	Phi-3 Mini 128K Instruct	51
10	Llama 3 8B Instruct	50
11	Mistral 7B Instruct (v0.2)	49
12	Yi-9B-200K	46
13	Llama 2 7B 32K Instruct	43
14	LongChat-7B-v1.5-32k	41
15	Llama 2 7B 32K	40

Interactive version: theaggregate.ai/benchmark?slug=babilong-niah-v0-4k · How It Works · Data refreshed daily, snapshot 2026-10-09.