Skip to content

Benchmark Scores

Public

Compare models on public benchmarks from Epoch AI, Artificial Analysis and Aider, against AnyRouter price and measured speed. Every chart names its source and fetch date.

Compare models

Pick a source and a benchmark, then 2–6 models. Only scores with a recorded source are shown.

1GPT-6 Astra2Claude Sonnet 5.53Gemini 3.8 Flash
GPQA Diamond
Top 15 of 111 models, higher is better.
  1. 1GPT-6 Astra95.8%
  2. 2Claude Sonnet 5.595.6%
  3. 3Gemini 3.8 Flash95.4%
  4. 4GPT-6.1 Sol95.4%
  5. 5Gemini 3.7 Flash94.8%
  6. 6GPT-6 Sol94.3%
  7. 7Gemini 3.1 Pro94.1%
  8. 8Grok 4.694.0%
  9. 9Claude Opus 5.594.0%
  10. 10Claude Opus 593.7%
  11. 11GPT-5.593.5%
  12. 12Kimi K393.5%
  13. 13GPT-5.6 Sol93.1%
  14. 14Grok 4.593.1%
  15. 15MiniMax M392.9%

Source: Epoch AI (CC-BY 4.0), fetched 2026-10-13; Artificial Analysis, fetched 2026-07-30

Side by side
Benchmarks at least two picked models report. Best per row in bold; — means no published score.
BenchmarkGPT-6 AstraClaude Sonnet 5.5Gemini 3.8 Flash
GPQA Diamond95.8%95.6%95.4%
ARC-AGI-295.0%—89.2%
SciCode56.5%61.0%56.6%
OTIS Mock AIME100.0%100.0%98.9%
CursorBench—55.5%39.6%
SimpleQA (verified)75.6%46.5%69.7%

Source: Epoch AI (CC-BY 4.0), fetched 2026-10-13

GPQA Diamond vs price
Blended AnyRouter list price per 1M tokens (3:1 input:output), log scale.

Source: Epoch AI (CC-BY 4.0), fetched 2026-10-13; Artificial Analysis, fetched 2026-07-30. Price: AnyRouter catalog; 9 free or unpriced models not plotted

GPQA Diamond vs speed
Output tokens per second measured on AnyRouter traffic.
No models have both values yet.

Source: Epoch AI (CC-BY 4.0), fetched 2026-10-13; Artificial Analysis, fetched 2026-07-30. Speed: AnyRouter network

Filter by Category

Top Models Overview

Top 30 by Intelligence Index
Artificial Analysis benchmark scores
reasoning Comparison
Average across 7 benchmarks
math Comparison
Average across 4 benchmarks
agentic Comparison
Average across 2 benchmarks
Model Performance
Benchmark score radar chart · 1 selected