Leaderboard

Market-Bench

Models write executable backtesters from plain-language descriptions of three strategies: scheduled trading on MSFT, KO/PEP pairs trading, and MSFT delta hedging. Outputs are compared with a reference implementation.

Best reported

443.2

grok-4

mean MAE

date not stated

Market-Bench paper v2 (2026-01-20)

mean MAE, lower is better. 10 models.

Updated date not stated

  1. grok-4443.2
  2. gpt-5.2969.4
  3. gemini-3-pro-preview1744.3
  4. gpt-5.1-codex-max4242.9
  5. deepseek-v3.24575.6
  6. claude-sonnet-4.55126.8
  7. claude-opus-4.56039.6
  8. command-a6562.1
  9. nova-premier-v17740.3
  10. llama-3.1-nemotron-ultra9674.2

Market-Bench paper v2 (2026-01-20)

mean MAE, lower is better. 10 models.

Updated date not stated

  1. 1grok-4443.2
  2. 2gpt-5.2969.4
  3. 3gemini-3-pro-preview1744.3
  4. 4gpt-5.1-codex-max4242.9
  5. 5deepseek-v3.24575.6
  6. 6claude-sonnet-4.55126.8
  7. 7claude-opus-4.56039.6
  8. 8command-a6562.1
  9. 9nova-premier-v17740.3
  10. 10llama-3.1-nemotron-ultra9674.2
Market-Bench leaderboard (mean MAE, lower is better)
1grok-4xAI443.2Defaultmarketbench.ai leaderboard
Official board, date not stated
2gpt-5.2OpenAI969.4Defaultmarketbench.ai leaderboard
Official board, date not stated
3gemini-3-pro-previewGoogle1744.3Defaultmarketbench.ai leaderboard
Official board, date not stated
4gpt-5.1-codex-maxOpenAI4242.9Defaultmarketbench.ai leaderboard
Official board, date not stated
5deepseek-v3.2DeepSeek4575.6Defaultmarketbench.ai leaderboard
Official board, date not stated
6claude-sonnet-4.5Anthropic5126.8Defaultmarketbench.ai leaderboard
Official board, date not stated
7claude-opus-4.5Anthropic6039.6Defaultmarketbench.ai leaderboard
Official board, date not stated
8command-aCohere6562.1Defaultmarketbench.ai leaderboard
Official board, date not stated
9nova-premier-v1Amazon7740.3Defaultmarketbench.ai leaderboard
Official board, date not stated
10llama-3.1-nemotron-ultraMeta9674.2Defaultmarketbench.ai leaderboard
Official board, date not stated

Full details for Market-Bench