Market-Bench: leaderboard

Market-Bench (2026) places LLM-driven retailer agents together in one simulated supply-chain market where they bid for stock in budget-constrained auctions, set prices and write slogans for persona-driven buyers; values are means over 10 runs of a 6-step simulation at temperature 0. This board is cumulative profit in simulation currency units, which can be negative; higher is better.

Metric: Cumulative profit (simulation currency units, signed; each agent starts with 22,500 in funds); all 20 models compete as retailer agents in one simulated supply-chain market (sealed-bid procurement auctions, retail pricing and slogans; 6 steps, temperature 0), mean over 10 runs; higher is better. Source: arxiv.org. Saturation forecast: Around April 2027. 19 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Pro36589#276
2Gemini 2.5 Flash26104#463
3O313368#237
4Claude Sonnet 4.510619#261
5GPT-4o7619#751
6Phi-47565#1931
7Qwen 2.5 VL 72B Instruct3402#841
8Market-Bench Llama 3.1 70B (checkpoint unspecified)2444
9QwenLong-L1-32B2242#1881
10Market-Bench Qwen2.5 32B (checkpoint unspecified)1815
11Market-Bench Gemma 3 27B (checkpoint unspecified)1481
12Qwen 2.5 VL 32B Instruct1409#1062
13ERNIE 4.5 300B A47B1360#1122
14Market-Bench InternLM2.5 20B (checkpoint unspecified)1154
15InternLM3-8B-Instruct1022#2703

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=market-bench · How It Works · Data refreshed daily, snapshot 2026-11-21.