Best AI for Trading and portfolios (November 2026)

Data refreshed 1 Nov 2026

Qwen3-235B-Think leads StockBench with 2.5 (final return %). InvestorAgent / GPT-4.1 leads Agent Market Arena (AMA) with 40.8 (cumulative return % (TSLA)). Qwen2.5-72B-Instruct leads InvestorBench with 46.2 (cumulative return %). Qwen3.6-Plus leads KTD-Fin with 85.3 (total return %). GPT-4.1 leads LiveTradeBench with 6.3 (cumulative return %). GLM-5.1 leads PortBench with 0.384 (CEPS (0-1)). MiMo-V2-Flash leads PolyBench with 17.6 (Confidence-Weighted Return (CWR) %). MiniMax-M2 leads AI-Trader with 9.6 (cumulative return %).

Benchmarks in this use case
11
With dated scores
8
Models with scores
67
Latest source date
9 Oct 2026

Trading and portfolios picks

The leader on each benchmark, never combined

  • Best on StockBench

    Qwen3-235B-ThinkQwen

    2.5 final return %

    Reported by the authors, 2 Mar 2026

  • Best on Agent Market Arena (AMA)

    InvestorAgent / GPT-4.1Maker not stated

    40.8 cumulative return % (TSLA)

    Reported by the authors, 30 Oct 2025

  • Best on InvestorBench

    Qwen2.5-72B-InstructQwen

    46.2 cumulative return %

    Reported by the authors, 24 Dec 2024

  • Best on KTD-Fin

    Qwen3.6-PlusQwen

    85.3 total return %

    Reported by the authors, 27 May 2026

  • Best on LiveTradeBench

    GPT-4.1OpenAI

    6.3 cumulative return %

    Reported by the authors, 5 Nov 2025

  • Best on PortBench

    GLM-5.1Zhipu AI

    0.384 CEPS (0-1)

    Reported by the authors, 9 Oct 2026

  • Best on PolyBench

    MiMo-V2-FlashXiaomi

    17.6 Confidence-Weighted Return (CWR) %

    Reported by the authors, 3 Apr 2026

  • Best on AI-Trader

    MiniMax-M2MiniMax

    9.6 cumulative return %

    Reported by the authors, 1 Dec 2025

Trading and portfolios leaderboard

StockBench arXiv v2 (2026-03-02)

final return %, higher is better. Top 5 of 11 models.

Updated 2 Mar 2026

  1. 1Qwen3-235B-Think2.5
  2. 2Qwen3-235B-Ins2.4
  3. 3GLM-4.52.3
  4. 4Claude-4-Sonnet2.2
  5. 5Qwen3-30B-Think2.1
StockBench leaderboard (final return %, higher is better)
1Qwen3-235B-ThinkQwen2.5composite rank 4Paper Table 2
Reported by the authors, 2 Mar 2026
2Qwen3-235B-InsQwen2.4composite rank 2Paper Table 2
Reported by the authors, 2 Mar 2026
3GLM-4.5Zhipu AI2.3composite rank 3Paper Table 2
Reported by the authors, 2 Mar 2026
4Claude-4-SonnetAnthropic2.2composite rank 7Paper Table 2
Reported by the authors, 2 Mar 2026
5Qwen3-30B-ThinkQwen2.1composite rank 6Paper Table 2
Reported by the authors, 2 Mar 2026
6Kimi k2Moonshot AI1.9composite rank 1 (z-score of return, drawdown, Sortino); avg of 3 runsPaper Table 2 (same as project leaderboard)
Reported by the authors, 2 Mar 2026
6OpenAI-O3OpenAI1.9composite rank 5Paper Table 2
Reported by the authors, 2 Mar 2026
8DeepSeek V3.1DeepSeek1.1composite rank 8Paper Table 2
Reported by the authors, 2 Mar 2026
9gpt-5OpenAI0.300composite rank 9Paper Table 2
Reported by the authors, 2 Mar 2026
10DeepSeek-V3DeepSeek0.200composite rank 11 (passive baseline is rank 12 at +0.4%)Paper Table 2
Reported by the authors, 2 Mar 2026
10Qwen3-CoderQwen0.200composite rank 10Paper Table 2
Reported by the authors, 2 Mar 2026

Agent Market Arena (AMA) arXiv v2 (2025-10-30); live board ongoing

cumulative return % (TSLA), higher is better. Top 5 of 10 models.

Updated 30 Oct 2025

  1. 1InvestorAgent / GPT-4.140.8
  2. 2InvestorAgent / GPT-4o33.0
  3. 3InvestorAgent / Vote29.1
  4. 4InvestorAgent / Claude-sonnet-428.9
  5. 5TradeAgent / Gemini-2.0-flash21.9
Agent Market Arena (AMA) leaderboard (cumulative return % (TSLA), higher is better)
1InvestorAgent / GPT-4.140.8live 2025-08-01 to 2025-09-30, TSLA; buy-and-hold 46.88Paper Table 1
Reported by the authors, 30 Oct 2025
2InvestorAgent / GPT-4o33.0live, TSLAPaper Table 1
Reported by the authors, 30 Oct 2025
3InvestorAgent / Vote29.1live, TSLAPaper Table 1
Reported by the authors, 30 Oct 2025
4InvestorAgent / Claude-sonnet-428.9live, TSLAPaper Table 1
Reported by the authors, 30 Oct 2025
5TradeAgent / Gemini-2.0-flash21.9live, TSLAPaper Table 1
Reported by the authors, 30 Oct 2025
6DeepFundAgent / GPT-4o13.9live, TSLAPaper Table 1
Reported by the authors, 30 Oct 2025
7InvestorAgent / Claude-3.5-haiku11.6live, TSLAPaper Table 1
Reported by the authors, 30 Oct 2025
8InvestorAgent / Gemini-2.0-flash10.6live, TSLAPaper Table 1
Reported by the authors, 30 Oct 2025
9DeepFundAgent / Vote8.6live, TSLAPaper Table 1
Reported by the authors, 30 Oct 2025
10TradeAgent / Vote-0.360live, TSLAPaper Table 1
Reported by the authors, 30 Oct 2025

InvestorBench ACL 2025 Long Papers (July 2025); arXiv v1 (2024-12-24)

cumulative return %, higher is better. Top 5 of 10 models.

Updated 24 Dec 2024

  1. 1Qwen2.5-72B-Instruct46.2
  2. 2GPT-443.7
  3. 3GPT-4o39.0
  4. 4Llama-3.1-70B-Instruct38.9
  5. 5Yi-1.5-34B-Chat38.0
InvestorBench leaderboard (cumulative return %, higher is better)
1Qwen2.5-72B-InstructQwen46.2stock task, average over 7 stocks, 2020-10-01 to 2021-05-06; buy-and-hold 34.099Paper Table 2
Reported by the authors, 24 Dec 2024
2GPT-4OpenAI43.7stock task, 7-stock averagePaper Table 2
Reported by the authors, 24 Dec 2024
3GPT-4oOpenAI39.0stock task, 7-stock averagePaper Table 2
Reported by the authors, 24 Dec 2024
4Llama-3.1-70B-InstructMeta38.9stock task, 7-stock averagePaper Table 2
Reported by the authors, 24 Dec 2024
5Yi-1.5-34B-ChatNot stated38.0stock task, 7-stock averagePaper Table 2
Reported by the authors, 24 Dec 2024
6Qwen-2.5-Instruct-7BQwen29.5stock task, 7-stock averagePaper Table 2
Reported by the authors, 24 Dec 2024
7DeepSeek-V2-Lite (15.7B)DeepSeek28.7stock task, 7-stock averagePaper Table 2
Reported by the authors, 24 Dec 2024
8DeepSeek-67B-ChatDeepSeek26.9stock task, 7-stock averagePaper Table 2
Reported by the authors, 24 Dec 2024
9Llama-3.1-8B-InstructMeta25.5stock task, 7-stock averagePaper Table 2
Reported by the authors, 24 Dec 2024
10GPT-o1-previewOpenAI25.1stock task, 7-stock averagePaper Table 2
Reported by the authors, 24 Dec 2024

KTD-Fin arXiv v1 (2026-05-27)

total return %, higher is better. Top 5 of 10 models.

Updated 27 May 2026

  1. 1Qwen3.6-Plus85.3
  2. 2GPT 5.561.3
  3. 3doubao-seed-2.061.0
  4. 4claude-opus-4-758.8
  5. 5minimax-m2.756.3
KTD-Fin leaderboard (total return %, higher is better)
1Qwen3.6-PlusQwen85.3blinded x open-research, 2024-01-01 to 2026-04-10; alpha rank 2; CSI300 buy-and-hold +36.92Paper Table 4
Reported by the authors, 27 May 2026
2GPT 5.5OpenAI61.3blinded x open-research; alpha rank 6Paper Table 4
Reported by the authors, 27 May 2026
3doubao-seed-2.0ByteDance61.0blinded x open-research; alpha rank 3Paper Table 4
Reported by the authors, 27 May 2026
4claude-opus-4-7Anthropic58.8blinded x open-research; alpha rank 1Paper Table 4
Reported by the authors, 27 May 2026
5minimax-m2.7MiniMax56.3blinded x open-research; alpha rank 4Paper Table 4
Reported by the authors, 27 May 2026
6step-3.5-flashStepFun21.9blinded x open-research; alpha rank 5Paper Table 4
Reported by the authors, 27 May 2026
7gemini-3.1-proGoogle8.3blinded x open-research; alpha rank 7Paper Table 4
Reported by the authors, 27 May 2026
8deepseek-v4-proDeepSeek1.4blinded x open-research; alpha rank 8Paper Table 4
Reported by the authors, 27 May 2026
9GLM-5.1Zhipu AI-3.5blinded x open-research; alpha rank 9Paper Table 4
Reported by the authors, 27 May 2026
10Kimi K2.6Moonshot AI-24.2blinded x open-research; alpha rank 10Paper Table 4
Reported by the authors, 27 May 2026

LiveTradeBench arXiv v1 (2025-11-05)

cumulative return %, higher is better. Top 5 of 10 models.

Updated 5 Nov 2025

  1. 1GPT-4.16.3
  2. 2GPT-o36.0
  3. 3gpt-55.3
  4. 4Qwen2.5-72B-Instruct5.2
  5. 5Llama4-Scout4.7
LiveTradeBench leaderboard (cumulative return %, higher is better)
1GPT-4.1OpenAI6.3US stock market, live 2025-08-18 to 2025-10-24 (50 trading days)Paper Table 2
Reported by the authors, 5 Nov 2025
2GPT-o3OpenAI6.0US stock market, livePaper Table 2
Reported by the authors, 5 Nov 2025
3gpt-5OpenAI5.3US stock market, livePaper Table 2
Reported by the authors, 5 Nov 2025
4Qwen2.5-72B-InstructQwen5.2US stock market, livePaper Table 2
Reported by the authors, 5 Nov 2025
5Llama4-ScoutMeta4.7US stock market, livePaper Table 2
Reported by the authors, 5 Nov 2025
6Llama4-MaverickMeta4.5US stock market, livePaper Table 2
Reported by the authors, 5 Nov 2025
7grok-4xAI4.3US stock market, livePaper Table 2
Reported by the authors, 5 Nov 2025
8Claude Opus 4Anthropic3.9US stock market, livePaper Table 2
Reported by the authors, 5 Nov 2025
9Claude Opus 4.1Anthropic3.7US stock market, livePaper Table 2
Reported by the authors, 5 Nov 2025
10Claude-Sonnet-3.7Anthropic3.6US stock market, livePaper Table 2
Reported by the authors, 5 Nov 2025

PortBench arXiv v6 (2026-10-09)

CEPS (0-1), higher is better. Top 5 of 10 models.

Updated 9 Oct 2026

  1. 1GLM-5.10.384
  2. 2Qwen3.6-Plus0.300
  3. 3Qwen3.7-Max0.292
  4. 4Kimi K2.60.274
  5. 5Qwen3.6-35B-A3B0.272
PortBench leaderboard (CEPS (0-1), higher is better)
1GLM-5.1Zhipu AI0.384pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
2Qwen3.6-PlusQwen0.300pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
3Qwen3.7-MaxQwen0.292pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
4Kimi K2.6Moonshot AI0.274pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
5Qwen3.6-35B-A3BQwen0.272pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
6DS-V4-ProDeepSeek0.266pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
7DB-2.0-ProNot stated0.263pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
8HY3-PreviewTencent0.258pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
9DB-2.0-LiteNot stated0.244pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
10DS-V4-FlashDeepSeek0.226pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026

PolyBench arXiv v1 (2026-04-03)

Confidence-Weighted Return (CWR) %, higher is better. Top 5 of 7 models.

Updated 3 Apr 2026

  1. 1MiMo-V2-Flash17.6
  2. 2Gemini-3-Flash6.2
  3. 3Trinity-Large-9.2
  4. 4deepseek-v3.2-11.4
  5. 5Grok-4.1-Fast-12.1
PolyBench leaderboard (Confidence-Weighted Return (CWR) %, higher is better)
1MiMo-V2-FlashXiaomi17.6base lot $10, 2026-02-06 to 2026-02-12Paper Table 5
Reported by the authors, 3 Apr 2026
2Gemini-3-FlashGoogle6.2base lot $10Paper Table 5
Reported by the authors, 3 Apr 2026
3Trinity-LargeNot stated-9.2base lot $10Paper Table 5
Reported by the authors, 3 Apr 2026
4deepseek-v3.2DeepSeek-11.4base lot $10Paper Table 5
Reported by the authors, 3 Apr 2026
5Grok-4.1-FastxAI-12.1base lot $10Paper Table 5
Reported by the authors, 3 Apr 2026
6GPT-OSS 120BOpenAI-17.8base lot $10Paper Table 5
Reported by the authors, 3 Apr 2026
7MiniMax-M2.5MiniMax-25.2base lot $10Paper Table 5
Reported by the authors, 3 Apr 2026

AI-Trader arXiv v1 (2025-12-01); repo updated 2026-06-11

cumulative return %, higher is better. Top 5 of 6 models.

Updated 1 Dec 2025

  1. 1MiniMax-M29.6
  2. 2DeepSeek V3.18.4
  3. 3Claude-3.7-Sonnet3.1
  4. 4gpt-51.6
  5. 5Qwen3-Max0.390
AI-Trader leaderboard (cumulative return %, higher is better)
1MiniMax-M2MiniMax9.6US market (Nasdaq-100, hourly), 2025-10-01 to 2025-11-07; QQQ baseline 1.87Paper Table 1
Reported by the authors, 1 Dec 2025
2DeepSeek V3.1DeepSeek8.4US market, hourlyPaper Table 1
Reported by the authors, 1 Dec 2025
3Claude-3.7-SonnetAnthropic3.1US market, hourlyPaper Table 1
Reported by the authors, 1 Dec 2025
4gpt-5OpenAI1.6US market, hourlyPaper Table 1
Reported by the authors, 1 Dec 2025
5Qwen3-MaxQwen0.390US market, hourlyPaper Table 1
Reported by the authors, 1 Dec 2025
6Gemini 2.5 FlashGoogle-0.060US market, hourlyPaper Table 1
Reported by the authors, 1 Dec 2025

Also in this use case

No dated scores yet