Atlas

Models

← All models

Yi 1.5 9B Chat

01.AI · 2024-05-13 · 8.8B parameters

Yi-1.5-9B-Chat is 01.AI's open-weight 9B chat checkpoint from the Yi-1.5 family, released on May 13, 2024. 01.AI describes Yi-1.5 as a continuation of Yi pretrained on a high-quality 500B-token corpus and fine-tuned on 3M diverse samples, with stronger coding, math, reasoning, and instruction-following performance. This exact checkpoint is configured for 4,096 tokens; the separately released Yi-1.5-9B-Chat-16K is a different artifact.

Benchmark scores

BenchmarkScore
BBH (Open LLM Leaderboard v2)55.6
BigCodeBench Complete42.4
BigCodeBench Instruct34.5
BigCodeBench-Hard Complete12.2
BigCodeBench-Hard Instruct10.1
Capability111.5
EQ-Bench v270.4
EQ-Bench v2 + MAGI-Hard Combined63.3
GPQA (Open LLM Leaderboard v2)33.5
IFEval (Open LLM Leaderboard v2)60.5
MAGI-Hard56.1
MATH Level 5 (Open LLM Leaderboard v2)22.6
MMLU-Pro (Open LLM Leaderboard v2)39.8
MuSR (Open LLM Leaderboard v2)42.6
ZeroEval33.0
ZeroEval CRUX44.8
ZeroEval GSM8K76.4
ZeroEval MATH Level 520.0
ZeroEval MMLU-Redux65.0
ZeroEval ZebraLogic2.3
Loading Atlas data…