The benchmark board

Every score from our Model Drops from the last 30 days on one board, since Oct 6, 2026: who leads, how fast the best scores move, and which new models jumped furthest past the last generation.

82benchmarks

48models scored

683scores logged

10Model Drops

Best model overall

Handy AI's weigh-in: every leaderboard on this page folded into one ranking.

Best score over time

Each dot is a model at its release. The line is the record holder at each point. Hollow dots are models without a drop of their own, placed where we first saw them cited.

AutomationBench

Best score rose 2.3 points, from 59.9% (Mistral Large 4) to 62.2% (MiniMax M3.1 Flash).

AA-LCR

Best score rose 3.6 points, from 82.7% (Claude Haiku 5.5) to 86.3% (Kimi K3.1 Pro).

Harvey Legal Agent Benchmark

Best score rose 7.9 points, from 15.8% (Mistral Large 4) to 23.7% (Kimi K3.1 Pro).

AutomationBench-AA

Best score rose 8.3 points, from 69.5% (Claude Opus 5.5) to 77.8% (Gemini 4 Carbon).

AA-Omniscience hallucination rate

Best score fell 28.3 points, from 51.3% (GPT-6 Astra) to 23% (Muse Wonder Full).

SWE-bench Pro

Best score rose 2 points, from 59.7% (GLM-5.3-Flash) to 61.7% (Qwen3.8).

SWE-Atlas-QnA

Best score rose 37.1 points, from 16.1% (Nemotron 3 Ultra) to 53.2% (Fern 0.1).

FrontierCode 2.0

Best score rose 27.3 points, from 15% (Mistral Large 4) to 42.3% (GPT-6.1 Terra).

Biggest generational jumps

A new model against the model it replaced, on the same chart in its launch post.

Claude Haiku 4.5Claude Haiku 5.5

OSWorld 2 (offline subset): 15.7% to 72.4%

+56.7

GPT-5.6 TerraGPT-6.1 Terra

AA-Omniscience hallucination rate: 88% to 38%

+50

DeepSeek V4.1 FlashDeepSeek V4.1 Flash 1017

Terminal-Bench 4.0: 31.2% to 42.3%

+11.1

Leaderboards

The latest reported score for every model on each benchmark. Tap a card's share button for an image of it.

Terminal-Bench 4.0

Multi-hour work in a real terminal.

Claude Sonnet 5.5max 70.6%

Claude Opus 5.5max 66.4%

Kimi K3.1 Promax 65.7%

GPT-6 Astramax 58.2%

Fable 5.1max 57.9%

Muse Wonder Fullmax 57.1%

24 more

GPT-6.1 Solmax 55.9%

Muse Wondermax 52%

Claude Opus 5 51.8%

GPT-6.1 Terramax 50.3%

Hy4 44.4%

DeepSeek V4.1 Flash 1017max 42.3%

GLM-5.3max 41.8%

GPT-5.6 Sol 39.9%

Claude Haiku 5.5max 39.2%

LongCat-2.5 38.5%

MiniMax M3.1 Flash 38.1%

DeepSeek V4.1 Flash 31.2%

Mistral Large 4 28.3%

Nanbeige4.3-33Bmax 23.8%

GPT-5.6 Terramax 21.5%

GPT-6 Luna 16.4%

Kimi K3 12.6%

DeepSeek V4 Pro 0813 10%

Qwen3.8max 6.3%

Fern 0.1 6.1%

Muse Glimmer-30Bmax 1.6%

Claude Haiku 4.5 0%

Gemma 4max 0%

Nemotron 3 Ultra 0%

30 models · 11 drops · GPT-6.1 Terra, Kimi K3.1 Pro, Muse Wonder, MiniMax M3.1 Flash, Tencent Hunyuan Hy4 launch post, Claude Haiku 5.5, Meituan LongCat-2.5 launch post, Mistral Large 4, Nanbeige4.3-33B model card, Mirendil, "Introducing Fern"

Sentiment Bench

How people received the model in its first week on X, Hacker News, in the news and among trusted reviewers (Handy AI).

Gemini 4 Carbon 85.8

DeepSeek V4.1 Flash 1017 83.9

Claude Haiku 5.5 71.4

GPT-6.1 Terra 69.2

Nemotron 3.5 Super VL 68.2

Kimi K3.1 Pro 67.5

4 more

MiniMax M3.1 Flash 61.1

Muse Spark Open 52.7

Mistral Large 4 49.4

Muse Wonder 37

10 models · 10 drops · Every model's sources, quotes and method

DeepSWE v1.1

Fixing issues in real repositories.

Muse Wonder Fullmax 79%

Muse Wondermax 78.8%

DeepSeek V4.1 Flash 1017max 77%

Kimi K3.1 Promax 75.5%

Claude Opus 5.5max 74.2%

DeepSeek V4.1 Flash 74.2%

18 more

GPT-6 Astramax 74.1%

Claude Opus 5 74%

MiniMax M3.1 Flash 73.7%

GPT-5.6 Sol 73%

Particle-1max 72.6%

Kimi K3 67.5%

Fable 5.1max 67.4%

GLM-5.3max 66.9%

Nanbeige4.3-33Bmax 65.5%

Mistral Large 4 62%

Muse Spark Openxhigh 60%

DeepSeek V4 Pro 0813 57%

Qwen3.8-2.4T-A95Bxhigh 57%

Fern 0.1 44.2%

Qwen3.8max 42.5%

Muse Glimmer-30Bmax 4.4%

Nemotron 3 Ultra 3.5%

Gemma 4max 2.7%

24 models · 9 drops · Muse Wonder, Kimi K3.1 Pro, MiniMax M3.1 Flash, Particle, "Introducing Particle-1", Nanbeige4.3-33B model card, Muse Spark Open, Mistral Large 4, Mirendil, "Introducing Fern"

GPQA Diamond

20 models compared.

GPT-6 Astramax 96.1%

Kimi K3.1 Promax 94.2%

Claude Opus 5.5max 93.9%

Fable 5.1max 93.7%

Qwen3.8-2.4T-A95Bxhigh 93.5%

Kimi K3max 93.5%

14 more

Qwen3.8-Flash-Next 92.7%

DeepSeek V4.1 Flash 1017max 92.2%

GLM-5.3max 91.7%

Gemma 4max 91.3%

Particle-1 90.4%

Nemotron 3.5 Super VL 89.5%

Nanbeige4.3-33Bmax 89.5%

Qwen3.8max 89.2%

Muse Spark Openxhigh 89%

Nemotron 3 Ultra 87%

Qwen3.5-122B-A10B 86.7%

Clover-35B-A3B-Preview 83.8%

Muse Glimmer-30Bmax 83.5%

LFM3-1.7B 63.4%

20 models · 7 drops · Kimi K3.1 Pro, Muse Spark Open, Nemotron 3.5 Super VL, Nanbeige4.3-33B model card, Particle, "Introducing Particle-1", DiffuSpace, "Clover-35B-A3B-Preview", Liquid AI, "Introducing LFM3"

AutomationBench

Real business automations (Zapier).

MiniMax M3.1 Flash 62.2%

DeepSeek V4.1 Flash 1017 60.5%

Mistral Large 4 59.9%

DeepSeek V4 Pro 0813 56.7%

Hy4 56.2%

DeepSeek V4.1 Flash 54.8%

10 more

Nanbeige4.3-33Bmax 54.2%

LongCat-2.5 51.8%

Claude Opus 5 50.3%

GLM-5.3 48.8%

Kimi K3 46.7%

GPT-5.6 Sol 45.8%

Qwen3.8max 38.3%

Muse Glimmer-30Bmax 15.2%

SenseNova-Loop-9B-Preview 13.7%

Gemma 4max 9.7%

16 models · 7 drops · MiniMax M3.1 Flash, Mistral Large 4, Tencent Hunyuan Hy4 launch post, Nanbeige4.3-33B model card, Meituan LongCat-2.5 launch post, SenseNova-Loop-9B-Preview model card

Terminal-Bench 2.1

13 models compared.

LongCat-2.5 88.4%

Kimi K3max 85%

GLM-5.3max 83.9%

Qwen3.8-2.4T-A95Bxhigh 82%

Nanbeige4.3-33Bmax 80.9%

Particle-1max 79.9%

7 more

Muse Spark Openxhigh 76.4%

Qwen3.8max 73%

Muse Glimmer-30Bmax 61.8%

Clover-35B-A3B-Preview 51.7%

Gemma 4max 47.2%

SenseNova-Loop-9B-Preview 40.5%

LFM3-1.7B 12.9%

13 models · 7 drops · Meituan LongCat-2.5 launch post, Muse Spark Open, Nanbeige4.3-33B model card, Particle, "Introducing Particle-1", DiffuSpace, "Clover-35B-A3B-Preview", SenseNova-Loop-9B-Preview model card, Liquid AI, "Introducing LFM3"

Artificial Analysis Intelligence Index

25 models compared.

Claude Opus 5.5 58

Claude Sonnet 5.5max 56

Muse Wonder Fullmax 55.8

Kimi K3.1 Pro 55

GPT-6 Astra 53

Fable 5.1 53

19 more

Gemini 4 Argonhigh 53

Muse Wondermax 52.4

GPT-6.1 Solmax 52

Claude Opus 5 50.8

Muse Spark 1.3 48.1

GPT-6.1 Terramax 48

MiMo-V2.6-Pro 46.3

DeepSeek V4.1 Flash 1017 45.1

GLM-5.3 44.8

Kimi K3 44

Claude Haiku 5.5max 43

MiniMax M3.1 Flash 42.7

GPT-5.6 Terramax 42

Qwen3.8-2.4T-A95B 39.9

DeepSeek V4.1 Flash 39.5

Muse Spark Open 38.9

Mistral Large 4 38.4

GPT-6 Lunamax 38

DeepSeek V4 Pro 0813 36

25 models · 6 drops · Kimi K3.1 Pro, GPT-6.1 Terra, Muse Wonder, DeepSeek V4.1 Flash 1017, Muse Spark Open, Mistral Large 4

Humanity's Last Exam (no tools)

Expert-level questions, no tools.

Claude Opus 5.5max 59.8%

Fable 5.1max 57.6%

Claude Sonnet 5.5 56.9%

Claude Opus 5 56.3%

GPT-6 Astramax 53.6%

Kimi K3.1 Promax 51.1%

11 more

Claude Haiku 5.5max 45.9%

GPT-5.6 Sol 44.5%

Kimi K3 43.5%

DeepSeek V4.1 Flash 1017 42.6%

Qwen3.8-Flash-Next 39.6%

DeepSeek V4.1 Flash 36.8%

Nemotron 3.5 Super VL 31.8%

Nemotron 3 Ultra 29.2%

Qwen3.5-122B-A10B 24.2%

Claude Haiku 4.5 10.2%

LFM3-1.7B 7.2%

17 models · 5 drops · Kimi K3.1 Pro, Claude Haiku 5.5, DeepSeek V4.1 Flash 1017, Nemotron 3.5 Super VL, Liquid AI, "Introducing LFM3"

ProgramBench

14 models compared.

Claude Opus 5.5max 65%

Muse Wonder Fullmax 59%

Fable 5.1max 53.5%

Muse Wondermax 50%

GPT-6 Astramax 50%

Claude Opus 5 37%

8 more

Kimi K3.1 Promax 31.5%

Hy4 30%

DeepSeek V4.1 Flash 1017 29.1%

LongCat-2.5 25.5%

GPT-5.6 Sol 23%

DeepSeek V4.1 Flash 20.3%

GLM-5.3max 19%

Kimi K3 17.5%

14 models · 5 drops · Kimi K3.1 Pro, Muse Wonder, DeepSeek V4.1 Flash 1017, Tencent Hunyuan Hy4 launch post, Meituan LongCat-2.5 launch post

GDPval-AA

Professional work products, graded head to head (Elo).

Claude Opus 5.5max 1866 Elo

Muse Wonder Fullmax 1857 Elo

Kimi K3.1 Promax 1846 Elo

Claude Sonnet 5.5 1838 Elo

Muse Wondermax 1770 Elo

Fable 5.1max 1754 Elo

10 more

DeepSeek V4.1 Flash 1017max 1746 Elo

GLM-5.3max 1650 Elo

Claude Haiku 5.5max 1620 Elo

Qwen3.8-2.4T-A95Bxhigh 1613 Elo

GPT-6 Astramax 1573 Elo

Kimi K3max 1537 Elo

GPT-6 Luna 1437 Elo

Mistral Large 4 1424 Elo

Muse Spark Openxhigh 1401 Elo

Claude Haiku 4.5 735 Elo

16 models · 4 drops · Kimi K3.1 Pro, Muse Wonder, Claude Haiku 5.5, Muse Spark Open

Agents' Last Exam

11 models compared.

Kimi K3.1 Promax 43.2%

Claude Opus 5.5max 42.6%

Hy4 41%

DeepSeek V4.1 Flash 1017 35.4%

GPT-6 Astramax 32.8%

Claude Opus 5 32.2%

5 more

DeepSeek V4.1 Flash 31.8%

GPT-5.6 Sol 30.6%

MiniMax M3.1 Flash 29.6%

Kimi K3 28.3%

GLM-5.3max 26.2%

11 models · 4 drops · Kimi K3.1 Pro, Tencent Hunyuan Hy4 launch post, DeepSeek V4.1 Flash 1017, MiniMax M3.1 Flash

SWE-bench Verified

8 models compared.

GPT-6 Astra 86%

GLM-5.3-Flash 79.1%

Nemotron 3 Ultra 73.9%

Clover-35B-A3B-Preview 73.6%

Fern 0.1 73.4%

Qwen3.5-122B-A10B 71.9%

2 more

Nemotron 3.5 Super VL 71.1%

LFM3-1.7B 43.4%

8 models · 4 drops · Mirendil, "Introducing Fern", Nemotron 3.5 Super VL, DiffuSpace, "Clover-35B-A3B-Preview", Liquid AI, "Introducing LFM3"

AA-LCR

17 models compared.

Kimi K3.1 Pro 86.3%

Nanbeige4.3-33Bmax 84%

Muse Wonder Fullmax 83.7%

Muse Glimmer-30Bmax 83.3%

GPT-6.1 Terramax 82.7%

DeepSeek V4.1 Flash 1017max 82.7%

11 more

Claude Haiku 5.5max 82.7%

Ling-3.1-flash-1019 82.3%

Qwen3.8max 82%

Muse Wondermax 80.7%

MiniMax M3.1 Flash 79.7%

Gemini 4 Carbonhigh 79.3%

Muse Spark Openxhigh 78.3%

Gemma 4max 72%

Clover-35B-A3B-Preview 72%

Nemotron 3.5 Super VL 69%

LFM3-1.7B 52.8%

17 models · 3 drops · Artificial Analysis (2026-11-04), Nanbeige4.3-33B model card, DiffuSpace, "Clover-35B-A3B-Preview", Liquid AI, "Introducing LFM3"

OSWorld 2

Using a computer: apps, files and the web.

Claude Sonnet 5.5max 83.9%

GPT-6 Astramax 80.9%

Hy4 78.9%

GPT-6.1 Solmax 78%

GPT-6.1 Terramax 74.2%

Claude Opus 5 70.6%

6 more

GPT-5.6 Sol 65.7%

MiniMax M3.1 Flash 61.7%

Kimi K3 58.3%

GPT-5.6 Terramax 57.1%

GPT-6 Lunamax 55.5%

DeepSeek V4.1 Flash 54.3%

12 models · 3 drops · GPT-6.1 Terra, Tencent Hunyuan Hy4 launch post, MiniMax M3.1 Flash

Harvey Legal Agent Benchmark

Legal work end to end.

Kimi K3.1 Promax 23.7%

DeepSeek V4.1 Flash 1017max 21%

Fable 5.1max 19.1%

Mistral Large 4 15.8%

Muse Wonder Fullmax 13.1%

Kimi K3 12.9%

5 more

Qwen3.8 Max 10.4%

Muse Wondermax 10%

GPT-6 Astramax 8.6%

GLM-5.3 8.3%

Claude Opus 5.5max 4.2%

11 models · 3 drops · Kimi K3.1 Pro, Mistral Large 4, Muse Wonder

BabyVision (with tools)

10 models compared.

Claude Opus 5.5max 97.7%

GPT-6 Astramax 96.6%

Kimi K3.1 Promax 95.4%

Fable 5.1max 95.1%

Claude Opus 5 94.1%

DeepSeek V4.1 Flash 1017 92.5%

4 more

DeepSeek V4.1 Flash 89.6%

Particle-1 89.2%

GPT-5.6 Sol 88.9%

Kimi K3 85.7%

10 models · 3 drops · Kimi K3.1 Pro, DeepSeek V4.1 Flash 1017, Particle, "Introducing Particle-1"

NL2Repo

10 models compared.

Claude Opus 5 75.3%

DeepSeek V4.1 Flash 1017 68.2%

DeepSeek V4.1 Flash 65.4%

MiniMax M3.1 Flash 61.2%

Kimi K3 58%

GLM-5.3 58%

4 more

GPT-5.6 Sol 56.8%

Nanbeige4.3-33Bmax 53%

Qwen3.8max 48.3%

Muse Glimmer-30Bmax 27.4%

10 models · 3 drops · MiniMax M3.1 Flash, Nanbeige4.3-33B model card

ZeroBench (with tools)

10 models compared.

Kimi K3.1 Promax 63%

Claude Opus 5.5max 61%

Fable 5.1max 59%

GPT-6 Astramax 58%

GPT-5.6 Sol 53%

Claude Opus 5 52%

4 more

DeepSeek V4.1 Flash 1017 51%

Muse Spark Openxhigh 50%

DeepSeek V4.1 Flash 49%

Kimi K3max 41%

10 models · 3 drops · Kimi K3.1 Pro, DeepSeek V4.1 Flash 1017, Muse Spark Open

Terminal-Bench Science

Scientific computing tasks in a terminal.

GPT-6 Astramax 68.1%

Claude Opus 5.5max 63.3%

GPT-6.1 Terramax 58.3%

GPT-6.1 Solmax 57%

Fable 5.1max 40%

Kimi K3.1 Promax 38.1%

3 more

Hy4 19.8%

DeepSeek V4.1 Flashmax 15.7%

GLM-5.3max 8.1%

9 models · 3 drops · GPT-6.1 Terra, Kimi K3.1 Pro, Tencent Hunyuan Hy4 launch post

SciCode

6 models compared.

Ling-3.1-flash-1019 57%

Nanbeige4.3-33Bmax 52.6%

Qwen3.8max 46.7%

Gemma 4max 45.5%

Muse Glimmer-30Bmax 44.9%

LFM3-1.7B 28.5%

6 models · 3 drops · Ling-3.1-flash-1019 model card, Nanbeige4.3-33B model card, Liquid AI, "Introducing LFM3"

Tau3-Bench

6 models compared.

LFM3-1.7B 68%

Clover-35B-A3B-Preview 65.3%

Nanbeige4.3-33Bmax 48.8%

Qwen3.8max 48.5%

Muse Glimmer-30Bmax 33.7%

Gemma 4max 18.6%

6 models · 3 drops · Liquid AI, "Introducing LFM3", DiffuSpace, "Clover-35B-A3B-Preview", Nanbeige4.3-33B model card

Terminal-Bench 4.0 (standard harness)

Terminal-Bench run by a third party on one shared harness.

Kimi K3.1 Pro 67.2%

Gemini 4 Carbonhigh 65.2%

Claude Opus 5.5max 59.6%

GPT-6 Astramax 59.1%

Gemini 4 Argonmax 57.1%

Muse Wonder Fullmax 57.1%

13 more

GPT-6.1 Terramax 55.6%

Fable 5.1max 52%

Muse Wondermax 52%

GLM-5.3 38.9%

DeepSeek V4.1 Flash 1017max 33.8%

Ling-3.1-flash-1019 32.8%

Claude Haiku 5.5max 32.8%

MiniMax M3.1 Flash 25.8%

Mistral Large 4high 24.7%

Kimi K3 17.2%

DeepSeek V4 Pro 0813 14.1%

Muse Spark Openxhigh 8.1%

Nemotron 3.5 Super VL 0%

19 models · 2 drops · Artificial Analysis (2026-11-04), Gemini 4 Carbon, Mistral Large 4

AutomationBench-AA

Business automations, run by Artificial Analysis.

Gemini 4 Carbonhigh 77.8%

Kimi K3.1 Pro 73.3%

MiniMax M3.1 Flash 72.5%

Muse Wonder Fullmax 71.3%

Claude Opus 5.5max 69.5%

GPT-6 Astramax 68.5%

10 more

Muse Wondermax 68.1%

Ling-3.1-flash-1019 67.5%

DeepSeek V4.1 Flash 1017max 66.6%

GLM-5.3max 62.2%

GPT-6.1 Terramax 60.2%

Fable 5.1max 59.4%

Mistral Large 4high 57.8%

Muse Spark Openxhigh 49.7%

Claude Haiku 5.5max 35.4%

Nemotron 3.5 Super VL 14.8%

16 models · 2 drops · Artificial Analysis (2026-11-04), Kimi K3.1 Pro

Chartography (with tools)

9 models compared.

GPT-6 Astramax 91%

Kimi K3.1 Promax 90.2%

Claude Opus 5.5max 89%

Fable 5.1max 88.4%

Claude Opus 5 84%

DeepSeek V4.1 Flash 1017 83%

3 more

GPT-5.6 Sol 79.9%

DeepSeek V4.1 Flash 78.9%

Kimi K3 68.1%

9 models · 2 drops · Kimi K3.1 Pro, DeepSeek V4.1 Flash 1017

OSWorld 2 (offline subset)

Using a computer, on the 82 OSWorld 2.1 tasks that need no internet.

Claude Opus 5.5max 87.2%

Claude Sonnet 5.5 83.9%

Gemini 4 Carbonmax 81.5%

GPT-6 Astramax 80.9%

Fable 5.1max 76.8%

Gemini 4 Argonmax 74.7%

3 more

Claude Haiku 5.5max 72.4%

GPT-6 Luna 48.9%

Claude Haiku 4.5 15.7%

9 models · 2 drops · Gemini 4 Carbon, Claude Haiku 5.5

AA-Omniscience hallucination rate

Lower is better.

Muse Wonder Fullmax 23%

Muse Wondermax 26.5%

GPT-6.1 Terra 38%

Claude Sonnet 5.5 47%

GPT-6 Astra 51.3%

Claude Opus 5.5 58.6%

2 more

Fable 5.1 72.6%

GPT-5.6 Terra 88%

8 models · 2 drops · Muse Wonder, GPT-6.1 Terra

SWE-bench Pro

8 models compared.

Qwen3.8max 61.7%

Qwen3.8-Flash-Next 60.4%

Nanbeige4.3-33Bmax 60.3%

GLM-5.3-Flash 59.7%

Nemotron 3.5 Super VL 51.8%

Gemma 4max 51.2%

2 more

Nemotron 3 Ultra 49.7%

Muse Glimmer-30Bmax 36.9%

8 models · 2 drops · Nanbeige4.3-33B model card, Nemotron 3.5 Super VL

SWE-Marathon

8 models compared.

Muse Wonder Fullxhigh 57.5%

Muse Wondermax 53.1%

Kimi K3.1 Promax 51.7%

Claude Opus 5.5max 46.3%

Fable 5.1max 45.6%

GPT-6 Astramax 42.5%

2 more

GLM-5.3max 42.5%

DeepSeek V4.1 Flash 1017max 40%

8 models · 2 drops · Muse Wonder, Kimi K3.1 Pro

Humanity's Last Exam (with tools)

Expert-level questions, with search and code.

Kimi K3.1 Promax 70%

Claude Opus 5.5max 67.7%

Fable 5.1max 65.6%

GLM-5.3max 62.5%

DeepSeek V4.1 Flash 1017max 58.8%

GPT-6 Astramax 57.2%

1 more

LFM3-1.7B 13.9%

7 models · 2 drops · Kimi K3.1 Pro, Liquid AI, "Introducing LFM3"

JobBench

7 models compared.

Muse Spark Openxhigh 61.8%

GLM-5.3max 61.4%

Kimi K3max 54.3%

Qwen3.8-2.4T-A95Bxhigh 52.7%

Nanbeige4.3-33Bmax 44.8%

Qwen3.8max 34.5%

1 more

Gemma 4max 9%

7 models · 2 drops · Muse Spark Open, Nanbeige4.3-33B model card

SWE-Atlas-QnA

6 models compared.

Fern 0.1 53.2%

Nanbeige4.3-33Bmax 50.8%

Qwen3.8max 41.9%

Muse Glimmer-30Bmax 17.7%

Nemotron 3 Ultra 16.1%

Gemma 4max 3.3%

6 models · 2 drops · Mirendil, "Introducing Fern", Nanbeige4.3-33B model card

EnterpriseOps-Gym

5 models compared.

Nanbeige4.3-33Bmax 41%

Qwen3.8max 40.2%

Muse Glimmer-30Bmax 30.5%

LFM3-1.7B 28.2%

Gemma 4max 27.3%

5 models · 2 drops · Nanbeige4.3-33B model card, Liquid AI, "Introducing LFM3"

IFBench

5 models compared.

Nemotron 3 Ultra 82.9%

LFM3-1.7B 73%

Nemotron 3.5 Super VL 70.9%

Qwen3.5-122B-A10B 67.6%

GLM-5.3-Flash 57.7%

5 models · 2 drops · Nemotron 3.5 Super VL, Liquid AI, "Introducing LFM3"

MCP Atlas

5 models compared.

Muse Spark Openxhigh 89.2%

Qwen3.8-2.4T-A95Bxhigh 84.5%

GLM-5.3max 84.2%

Kimi K3max 82.3%

LFM3-1.7B 32.6%

5 models · 2 drops · Muse Spark Open, Liquid AI, "Introducing LFM3"

BrowseComp

4 models compared.

Kimi K3.1 Promax 92.1%

LongCat-2.5 91.9%

GPT-6 Astramax 91.5%

Claude Opus 5.5max 90.8%

4 models · 2 drops · Kimi K3.1 Pro, Meituan LongCat-2.5 launch post

From a single launch

Artificial Analysis Intelligence Index v5.0

18 models compared.

Gemini 4 Carbonhigh 60

Claude Opus 5.5 58

GPT-6 Astra 57

GPT-6.1 Sol 55

Claude Sonnet 5.5 54

GPT-6.1 Terramax 51.3

12 more

Muse Wonder Fullmax 50.1

Gemini 4 Argon 50

Fable 5.1 49

Kimi K3.1 Pro 46.5

Muse Wondermax 43.6

Claude Haiku 5.5max 36.5

DeepSeek V4.1 Flash 1017max 32.6

Ling-3.1-flash-1019 29.6

MiniMax M3.1 Flash 29.1

Mistral Large 4high 26.1

Muse Spark Openxhigh 24.9

Nemotron 3.5 Super VL 13.6

18 models · 1 drop · Artificial Analysis (2026-11-04), Gemini 4 Carbon

AA-Briefcase v1.1

16 models compared.

Claude Opus 5.5max 1803 Elo

Muse Wonder Fullmax 1729 Elo

Kimi K3.1 Pro 1675 Elo

Fable 5.1max 1672 Elo

Muse Wondermax 1659 Elo

Gemini 4 Carbonhigh 1610 Elo

10 more

Claude Haiku 5.5max 1574 Elo

DeepSeek V4.1 Flash 1017max 1570 Elo

GPT-6 Astramax 1569 Elo

Ling-3.1-flash-1019 1522 Elo

GLM-5.3max 1507 Elo

MiniMax M3.1 Flash 1475 Elo

GPT-6.1 Terramax 1401 Elo

Mistral Large 4high 1348 Elo

Muse Spark Openxhigh 1278 Elo

Nemotron 3.5 Super VL 999 Elo

16 models · 1 drop · Kimi K3.1 Pro, Artificial Analysis (2026-11-04)

Vals Index v2.2

14 models compared.

Gemini 4 Carbon 70.5%

Claude Opus 5.5max 70.2%

Gemini 4 Argonmax 69.2%

Muse Wonder Full 66.8%

GPT-6 Astramax 65%

Muse Wonder 64.9%

8 more

Kimi K3.1 Pro 62.5%

DeepSeek V4.1 Flash 1017 61.3%

GPT-6.1 Terra 60.5%

Claude Haiku 5.5 58.7%

Mistral Large 4 56.5%

MiniMax M3.1 Flash 53.8%

Muse Spark Open 48.6%

Nemotron 3.5 Super VL 29.8%

14 models · 1 drop · Vals Index v2.2 (2026-11-04), Gemini 4 Carbon

FrontierCode 2.0

Real coding tasks in open-source repos (Cognition), run by Artificial Analysis.

GPT-6.1 Terramax 42.3%

Gemini 4 Carbonhigh 40.6%

Muse Wonder Fullmax 37.8%

Claude Haiku 5.5max 34.9%

Hy4 31%

DeepSeek V4.1 Flash 1017max 28.5%

7 more

Muse Wondermax 27.9%

Kimi K3.1 Pro 25.5%

MiniMax M3.1 Flash 20.8%

Ling-3.1-flash-1019 17%

Mistral Large 4high 15%

Muse Spark Openxhigh 7.6%

Nemotron 3.5 Super VL 6.1%

13 models · 1 drop · Artificial Analysis (2026-11-04), Tencent Hunyuan Hy4 launch post

Artificial Analysis Cyber Index

8 models compared.

Kimi K3.1 Pro 64.1

DeepSeek V4.1 Flash 1017max 56.8

Mistral Large 4high 50.9

GLM-5.3-Flash 50

Kimi K3 41

GLM-5.3 36

2 more

GPT-6 Astra 33

Claude Opus 5.5 29

8 models · 1 drop · Artificial Analysis (2026-11-04), Mistral Large 4

Agents' Last Exam v1

6 models compared.

Claude Opus 5.5max 63.2%

GPT-6.1 Terramax 60%

GPT-6 Astramax 59.3%

GPT-6.1 Solmax 58.6%

GPT-6 Lunamax 50.9%

GPT-5.6 Terramax 50.7%

6 models · 1 drop · GPT-6.1 Terra

Apex Bench v1.1

6 models compared.

DeepSeek V4.1 Flash 1017 72.1%

Claude Opus 5 65.8%

DeepSeek V4.1 Flash 64.6%

GLM-5.3 56.6%

GPT-5.6 Sol 54.3%

Kimi K3 47.8%

6 models · 1 drop · DeepSeek V4.1 Flash 1017

APEX-Agents

6 models compared.

Claude Opus 5.5max 73.5%

Fable 5.1max 68.6%

Kimi K3.1 Promax 67.2%

GPT-6 Astramax 64.7%

DeepSeek V4.1 Flash 1017max 63.4%

GLM-5.3max 56.6%

6 models · 1 drop · Kimi K3.1 Pro

BankerToolBench

6 models compared.

MiniMax M3.1 Flash 76.2

Claude Opus 5 76.2

GPT-5.6 Sol 74.6

GLM-5.3 74.5

DeepSeek V4.1 Flash 73.7

Kimi K3 69.1

6 models · 1 drop · MiniMax M3.1 Flash

Frontier Finance

6 models compared.

Fable 5.1max 60.8%

Kimi K3.1 Promax 58.6%

GLM-5.3max 57.5%

Claude Opus 5.5max 56.7%

DeepSeek V4.1 Flash 1017max 54.5%

GPT-6 Astramax 45%

6 models · 1 drop · Kimi K3.1 Pro

GDP.pdf

Work products built from PDFs (Surge AI).

Kimi K3.1 Promax 34.6%

GPT-6 Astramax 34.2%

Claude Opus 5.5max 30.6%

Fable 5.1max 27.6%

DeepSeek V4.1 Flash 1017max 17.2%

GLM-5.3max 11.2%

6 models · 1 drop · Kimi K3.1 Pro

OfficeQA Pro v2

6 models compared.

Kimi K3.1 Promax 64.4%

Claude Opus 5.5max 62.2%

Fable 5.1max 54.5%

GPT-6 Astramax 54.4%

DeepSeek V4.1 Flash 1017max 40.4%

GLM-5.3max 31.1%

6 models · 1 drop · Kimi K3.1 Pro

Agents' Last Exam (CLI)

5 models compared.

Kimi K3.1 Promax 44.2%

Claude Opus 5.5max 40.7%

GPT-6 Astramax 38.8%

DeepSeek V4.1 Flash 1017max 36%

GLM-5.3max 29.4%

5 models · 1 drop · Kimi K3.1 Pro

Artificial Analysis Coding Agent Index v2.0

5 models compared.

Claude Opus 5.5max 62

Claude Sonnet 5.5max 59

GPT-6.1 Solmax 54

GPT-6.1 Terramax 48

GPT-6 Lunamax 31

5 models · 1 drop · GPT-6.1 Terra

BenchCAD (with tools)

5 models compared.

Claude Opus 5.5max 96.2%

GPT-6 Astramax 95.9%

Kimi K3.1 Promax 92.7%

Fable 5.1max 92.6%

DeepSeek V4.1 Flash 1017max 81.6%

5 models · 1 drop · Kimi K3.1 Pro

CyberGym-E2E

5 models compared.

Mistral Large 4 82%

MiMo-V2.6-Pro 79%

Grok 4.7 74%

Kimi K3 58%

GLM-5.3 29%

5 models · 1 drop · Mistral Large 4

ExploitBench

5 models compared.

GPT-6 Astramax 100%

Claude Opus 5.5max 91%

Fable 5.1max 83%

Kimi K3.1 Promax 74.7%

GLM-5.3max 54.4%

5 models · 1 drop · Kimi K3.1 Pro

Factual error rate

Lower is better.

GPT-6 Astramax 4.2%

GPT-6.1 Terramax 4.3%

GPT-6.1 Solmax 4.4%

GPT-6 Lunamax 18.4%

GPT-5.6 Terramax 22.3%

5 models · 1 drop · GPT-6.1 Terra

FrontierSWE v2

Frontier software engineering tasks.

GPT-6 Astramax 65.5%

Claude Opus 5.5max 62.3%

Gemini 4 Carbonmax 61.2%

Fable 5.1max 56.3%

Gemini 4 Argonmax 55%

5 models · 1 drop · Gemini 4 Carbon

MemoryBench

5 models compared.

Muse Wondermax 85.6%

Muse Wonder Fullmax 85.5%

Claude Opus 5.5max 81.5%

GPT-6 Astramax 81.4%

Fable 5.1max 79%

5 models · 1 drop · Muse Wonder

MMMU-Pro

5 models compared.

Claude Opus 5.5max 87.7%

GPT-6 Astramax 86.9%

Fable 5.1max 86.3%

Kimi K3.1 Promax 84.1%

DeepSeek V4.1 Flash 1017max 80.6%

5 models · 1 drop · Kimi K3.1 Pro

MMMU-Pro Vision

5 models compared.

GPT-6 Astramax 86.7%

GPT-6.1 Solmax 86.3%

GPT-6.1 Terramax 85.5%

GPT-5.6 Terramax 81.7%

GPT-6 Lunamax 79.9%

5 models · 1 drop · GPT-6.1 Terra

OSWorld

5 models compared.

Claude Opus 5.5max 81.8%

Fable 5.1max 80.7%

Kimi K3.1 Promax 79.6%

GPT-6 Astramax 75.8%

DeepSeek V4.1 Flash 1017max 69.3%

5 models · 1 drop · Kimi K3.1 Pro

Personal Superintelligence Index

5 models compared.

Muse Wonder Fullmax 59.1

Claude Opus 5.5max 55.2

GPT-6 Astramax 54.5

Muse Wondermax 54.3

Fable 5.1max 49.9

5 models · 1 drop · Muse Wonder

SWE-sweep

5 models compared.

Gemini 4 Carbonmax 9.6%

GPT-6 Astramax 9.5%

Gemini 4 Argonmax 4.3%

Claude Opus 5.5max 3.6%

Fable 5.1max 2.9%

5 models · 1 drop · Gemini 4 Carbon

Agents' Last Exam v1.1

4 models compared.

Gemini 4 Carbonmax 77.4%

GPT-6 Astramax 73.1%

Claude Opus 5.5max 71.6%

Gemini 4 Argonmax 69%

4 models · 1 drop · Gemini 4 Carbon

AIME 2026

4 models compared.

Nanbeige4.3-33Bmax 95.2%

Muse Glimmer-30Bmax 94.7%

Qwen3.8max 94.5%

Gemma 4max 89.2%

4 models · 1 drop · Nanbeige4.3-33B model card

Chartography

4 models compared.

Claude Sonnet 5.5 61.6%

Claude Haiku 5.5max 46.4%

GPT-6 Luna 29.1%

Claude Haiku 4.5 6.4%

4 models · 1 drop · Claude Haiku 5.5

CharXiv Reasoning

4 models compared.

Qwen3.5-122B-A10B 83.5%

GLM-5.3-Flash 81.9%

Nemotron 3.5 Super VL 80.8%

Inkling Small 77.8%

4 models · 1 drop · Nemotron 3.5 Super VL

Claw-Eval pass^3

4 models compared.

Nanbeige4.3-33Bmax 78.3%

Qwen3.8max 75.2%

Muse Glimmer-30Bmax 57.3%

Gemma 4max 44.2%

4 models · 1 drop · Nanbeige4.3-33B model card

CyberGym

4 models compared.

GLM-5.3 84.5%

GPT-5.6 Sol 83.6%

MiniMax M3.1 Flash 82.2%

Kimi K3 80%

4 models · 1 drop · MiniMax M3.1 Flash

DeepPlanning

4 models compared.

Nanbeige4.3-33Bmax 63.3%

Qwen3.8max 49.2%

Muse Glimmer-30Bmax 21.4%

Gemma 4max 19.4%

4 models · 1 drop · Nanbeige4.3-33B model card

Finance Agent v2

4 models compared.

GLM-5.3 55.8%

Mistral Large 4 54.7%

GPT-6 Astra 53.5%

Kimi K3 53.1%

4 models · 1 drop · Mistral Large 4

GDPval rubrics

4 models compared.

Qwen3.8max 86.4%

Nanbeige4.3-33Bmax 84.8%

Muse Glimmer-30Bmax 75.8%

Gemma 4max 71.4%

4 models · 1 drop · Nanbeige4.3-33B model card

HLE-Diamond (no tools)

4 models compared.

Muse Spark Openxhigh 26.1%

Kimi K3max 22.2%

Qwen3.8-2.4T-A95Bxhigh 17.3%

GLM-5.3max 16.4%

4 models · 1 drop · Muse Spark Open

SkillsBench

4 models compared.

Nanbeige4.3-33Bmax 61.1%

Qwen3.8max 56.4%

Muse Glimmer-30Bmax 39.7%

Gemma 4max 28.9%

4 models · 1 drop · Nanbeige4.3-33B model card

Toolathlon-Verified

4 models compared.

Nanbeige4.3-33Bmax 70.4%

Qwen3.8max 68.5%

Muse Glimmer-30Bmax 47.2%

Gemma 4max 27.8%

4 models · 1 drop · Nanbeige4.3-33B model card

Vals Index

4 models compared.

GLM-5.3 53.5%

Kimi K3 50.3%

Mistral Large 4 48.1%

DeepSeek V4 Pro 0813 47.6%

4 models · 1 drop · Mistral Large 4

τ²-bench Telecom

4 models compared.

GLM-5.3-Flash 98.5%

Qwen3.5-122B-A10B 87.2%

Nemotron 3 Ultra 83%

Nemotron 3.5 Super VL 73.2%

4 models · 1 drop · Nemotron 3.5 Super VL

Artificial Analysis Coding Agent Index

3 models compared.

Claude Opus 5.5 62

Gemini 4 Carbon 59

Gemini 4 Argon 53

3 models · 1 drop · Gemini 4 Carbon

Dense200

3 models compared.

Mistral Large 4 42%

GPT-6 Astra 41.5%

Kimi K3 28.9%

3 models · 1 drop · Mistral Large 4

FrontierCode v1.1

Hard, long coding tasks.

Claude Sonnet 5.5xhigh 52.1%

Claude Haiku 5.5max 46.4%

GPT-6 Luna 42.4%

3 models · 1 drop · Claude Haiku 5.5

SWE-bench Pro v2 Hard

3 models compared.

Claude Opus 5.5 100%

Fern 0.1 76.5%

Nemotron 3 Ultra 31.4%

3 models · 1 drop · Mirendil, "Introducing Fern"

AA-Omniscience

2 models compared.

Gemini 4 Carbon 54

Claude Opus 5.5 48

2 models · 1 drop · Gemini 4 Carbon

Hermes Bench

2 models compared.

Nemotron 3.5 Super VL 63.4%

Nemotron 3 Ultra 56.3%

2 models · 1 drop · Nemotron 3.5 Super VL

CritPt v2 Hard

Unpublished research-level physics problems, hard subset, run by Artificial Analysis.

Gemini 4 Carbonhigh 69%

GPT-6.1 Terramax 68%

Muse Wonder Fullmax 56.5%

Muse Wondermax 50.5%

Kimi K3.1 Pro 49.5%

Claude Haiku 5.5max 43.5%

6 more

Ling-3.1-flash-1019 38.5%

MiniMax M3.1 Flash 32.5%

Muse Spark Openxhigh 29.5%

Mistral Large 4high 27%

DeepSeek V4.1 Flash 1017max 25%

Nemotron 3.5 Super VL 13.5%

12 models · 0 drops · Artificial Analysis (2026-11-04)

Terminal-Bench Science (standard harness)

Terminal-Bench Science run by a third party on one shared harness.

Gemini 4 Carbonhigh 64.8%

GPT-6.1 Terramax 57.6%

Muse Wonder Fullmax 37.1%

Kimi K3.1 Pro 34.3%

Muse Wondermax 25.7%

Claude Haiku 5.5max 20%

6 more

DeepSeek V4.1 Flash 1017max 12.9%

Mistral Large 4high 4.3%

Ling-3.1-flash-1019 4.3%

MiniMax M3.1 Flash 3.3%

Muse Spark Openxhigh 1.9%

Nemotron 3.5 Super VL 0%

12 models · 0 drops · Artificial Analysis (2026-11-04)

Release pace

Every Model Drop by month, marked with its maker.

17 Oct 2026

2 Nov 2026

How this board works

Scores come from the benchmark section of each handyai.news Model Drop, mostly the makers' launch tables. When drops disagree, the newest drop wins, since later launches re-run rivals on the current version of a benchmark; the model's own launch number is kept as launch_score. Settings (effort level, harness) vary between makers, so compare within a board with care. Models without their own drop are dated by the first drop that cites them.

The data is open: /api/benchmarks (JSON), /benchmarks.md (Markdown), one benchmark at /api/benchmarks/{benchmark}, the best model overall at /api/benchmarks/weigh-in (or /weigh-in/{type} for a topic), a PNG of any card by adding .png, and the list_benchmarks, get_benchmark and get_weigh_in tools on our MCP server.