Machine Trading 0.11.0 - Market Making SPCX: leaderboard
Machine Trading release0.11.0: coding agents implement WebAssembly trading bots, graded on72hidden Solana market replays per task. Exact native Claude API identities, Claude Code2.1.312, five successful graded trials/model/task. Skill is each trial total PnL divided by the reference bot total PnL on the same sessions; board is the arithmetic mean of five trial skills. Signed and unbounded: losses can be negative and 1matches reference. Do not substitute reward(PnL/bound), win rate, market-making versus flash-crash tasks, or newer Hermes harness releases. Native effort/token budget is not reported; no guessed reasoning setting. Historical0.11.0rescorelast-bit differences do not replace the retained reported trial values. Native task: market-making-spcx
Metric: Mean total-PnL skill ratio (5 runs; reference = 1). Source: huggingface.co. Saturation forecast: Around April 2028. 4 models tracked. Updated 14 Nov 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5.2 | 12.04 | #1 |
| 2 | Claude Opus 5.5 | 1.22 | #5 |
| 3 | Claude Fable 5.1 | -0.9365 | #15 |
| 4 | Claude Sonnet 5.5 | -1.63 | #19 |
No result here: #2 Gemini 4 Carbon, #6 GPT-6 Astra, #10 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=machine-trading-0-11-0-market-making-spcx · How It Works · Data refreshed daily, snapshot 2026-11-16.