Atlas

Benchmarks

← All benchmarks

Market-Bench - Single-Stock Scheduled Execution (Mean of 5)

Professional Work · 2025-12-13

The easy Market-Bench strategy asks a model to implement scheduled buying and selling of MSFT against randomized market-by-price L10 liquidity while correctly reserving consumed synthetic-book volume. This definition reports mean absolute error averaged over five generated backtester runs.

Top models (lower is better)

ModelScore
Gemini 3 Pro Preview14.8
Sonnet 4.516.4
Opus 4.5111
Grok 4122
DeepSeek-V3.2139
Command A185
Nova Premier242
Mistral Large 3 675B Instruct 2512362
GPT-5.1-Codex-Max845
GPT-5.21546
Qwen3-Max (2025-09-23)2016
Llama 3.1 Nemotron Ultra 253B V12511
Llama 4 Maverick4138
Loading Atlas data…