Atlas

Models

← All models

Claude 2.1

Anthropic · 2023-11-21

Claude 2.1 is Anthropic's November 2023 update to Claude 2, increasing the supported context window from 100K to the full 200K tokens and improving long-context reliability. Anthropic also reported fewer false statements and incorrect long-document answers, and introduced system prompts and beta API tool use with the release.

Benchmark scores

BenchmarkScore
AlpacaEval 2.0 (length-controlled win rate)25.3
AlpacaEval 2.0 (raw win rate)15.7
Artificial Analysis Coding Index14.0
Artificial Analysis Progress Index v1116
Capability109.2
DTBench51.0
Epoch Capabilities Index (ECI)119
EQ-Bench v274.0
ForecastBench Baseline Dataset Brier Index52.7
ForecastBench Baseline Market Brier Index53.7
ForecastBench Baseline Overall Brier Index53.2
ForecastBench Preliminary Dataset Brier Index52.7
ForecastBench Tournament Dataset Brier Index52.7
ForecastBench Tournament Market Brier Index65.8
ForecastBench Tournament Overall Brier Index58.7
GPQA Diamond33.0
GPQA Diamond (Artificial Analysis avg@5)31.9
Humanity's Last Exam (Text-Only)3.9
LiveCodeBench (Artificial Analysis)19.5
MATH-500 (Artificial Analysis source)37.4
MMLU Pro49.5
OTIS Mock AIME 2024-2025 (Epoch legacy model grader)1.9
SciCode (Artificial Analysis avg@3)18.4
WeirdML (v2)7.1
Loading Atlas data…