Atlas

Models

← All models

OpenChat 3.5 1210

OpenChat · 2023-12-18 · 7.2B parameters

OpenChat 3.5 1210 is a 7B open-weight chat model aligned with C-RLFT on mixed-quality instruction data. The 1210 release adds separate coding and generalist modes plus mathematical-reasoning improvements over OpenChat 3.5.

Benchmark scores

BenchmarkScore
Artificial Analysis Intelligence Index3.0
BBH (Open LLM Leaderboard v2)45.4
Capability104.8
Creative Writing v257.3
EQ-Bench v272.5
EQ-Bench v2 + MAGI-Hard Combined55.7
GPQA (Open LLM Leaderboard v2)30.1
GPQA Diamond23.0
Humanity's Last Exam (Text-Only)4.8
IFEval (Open LLM Leaderboard v2)60.4
LiveCodeBench (Artificial Analysis)11.5
MAGI-Hard38.8
MATH Level 5 (Open LLM Leaderboard v2)7.9
MATH-500 (Artificial Analysis source)30.7
MMLU Pro31.0
MMLU-Pro (Open LLM Leaderboard v2)31.4
MuSR (Open LLM Leaderboard v2)44.1
Loading Atlas data…