Last refreshed 2026-10-20 · 117 models · 40 tasks · graded by code, not by a model

Results: 117 models tested, 103 finished the full 40.

Every model below faced the same 40 revenue-operations jobs and the same fixed grading code. Ranked by mean score. 14 models stopped short of the full set; their rows sit outside the ranking with counts shown.

models tested
117
full coverage
103
partial runs
14
01 · TOP 15

Top 15 of 103 full-coverage models

Mean score across all 40 tasks. Each model name opens its graded answers in the reviewer.

#ModelMeanPerfectAvg latencyModalities
1meta/muse-wonder0.995839/4028.0stextimagefilevideoaudio
2deepseek/deepseek-pro-latest0.989638/40176.9stext
3elorian/glimpse0.989638/40125.3stextimagevideo
4fireworks/ember-10.988738/40124.8stextimage
5meta/muse-spark-1.10.986437/4040.2stextimagefilevideo
6moonshotai/kimi-k30.985437/40102.1stextimagevideo
7z-ai/glm-5.3-flashx0.983137/40114.6stextimagevideo
8unbiased/pareto-26.90.981937/4093.4stextimage
9scaledown/summarize0.981236/4073.7stextimagefile
10stepfun/step-5-preview0.981036/4092.6stextimagevideo
11google/gemini-3.8-flash frontier0.979636/40117.1stextimagefilevideoaudio
12qwen/qwen3.8-max-09020.979637/40159.5stextimagevideo
13meta/muse-spark-1.30.978336/4043.1stextimagefilevideo
14xiaomi/mimo-v2.5-pro0.977936/40119.1stext
15meta/muse-spark-open0.977536/40115.3stextimagefilevideoaudio

Scores compare within this run only. Frontier baselines carry a frontier tag. Modalities in means what you can send the model: text, images, files, video, or audio. Modalities out means what it can send back: text for 116 of 117 models, plus images for google/gemini-3-pro-image. Bench tasks feed text extracts, so modalities do not change the score. They are here so you know what each model can handle in real work.

02 · FULL BOARD

All 103 full-coverage models

Ranked by mean score, best first. Model names link to graded answers in the reviewer.

#ModelMeanPerfectAvg latencyModalities
1meta/muse-wonder0.995839/4028.0stextimagefilevideoaudio
2deepseek/deepseek-pro-latest0.989638/40176.9stext
3elorian/glimpse0.989638/40125.3stextimagevideo
4fireworks/ember-10.988738/40124.8stextimage
5meta/muse-spark-1.10.986437/4040.2stextimagefilevideo
6moonshotai/kimi-k30.985437/40102.1stextimagevideo
7z-ai/glm-5.3-flashx0.983137/40114.6stextimagevideo
8unbiased/pareto-26.90.981937/4093.4stextimage
9scaledown/summarize0.981236/4073.7stextimagefile
10stepfun/step-5-preview0.981036/4092.6stextimagevideo
11google/gemini-3.8-flash frontier0.979636/40117.1stextimagefilevideoaudio
12qwen/qwen3.8-max-09020.979637/40159.5stextimagevideo
13meta/muse-spark-1.30.978336/4043.1stextimagefilevideo
14xiaomi/mimo-v2.5-pro0.977936/40119.1stext
15meta/muse-spark-open0.977536/40115.3stextimagefilevideoaudio
16tencent/hy3-preview0.976835/4082.2stext
17anthropic/claude-opus-5 frontier0.976237/40106.9stextimagefile
18qwen/qwen3.8-flash0.974236/40111.3stextimagevideo
19meta/muse-spark-1.20.973335/4062.3stextimagefilevideo
20unbiased/pareto0.973336/4067.3stext
21z-ai/glm-5.10.972934/40178.4stext
22z-ai/glm-5.3-flash0.972135/40400.9stextimagevideo
23qwen/qwen3.8-omni-flash0.970034/40184.7stextimagevideoaudio
24z-ai/glm-5.30.969634/40301.3stext
25unbiased/pareto-26.100.967935/4084.2stextimage
26tencent/hy30.967136/401611.3stext
27anthropic/claude-opus-4.5 frontier0.966233/4090.3stextimagefile
28~unbiased/pareto-latest0.965834/40108.4stextimage
29openai/gpt-5.5-pro frontier0.965634/40183.4stextimagefile
30deepseek/deepseek-flash-latest0.965434/40151.4stextimage
31qwen/qwen3.8-2.4t-a95b0.965034/40259.1stext
32google/gemini-4-argon frontier0.964634/40139.2stextimagefilevideoaudio
33anthropic/claude-sonnet-5.5 frontier0.964334/4048.4stextimagefile
34deepseek/deepseek-v4-pro0.964234/40327.8stext
35xiaomi/mimo-v2.6-pro-ultraspeed0.964235/4048.3stextimagevideoaudio
36anthropic/claude-opus-5.5 frontier0.962934/4063.2stextimagefile
37google/gemini-3.7-flash frontier0.962532/4083.5stextimagefilevideoaudio
38qwen/qwen3.8-27b0.962134/40213.1stextimagevideo
39sakana/fugu-ultra0.961735/40410.2stextimage
40qwen/qwen3.8-max-prime0.961734/40108.8stextimagevideo
41xingchen-agi/xing4.0-29b-a4b:free0.961734/40145.0stext
42meituan/longcat-2.00.960433/40245.4stext
43anthropic/claude-opus-4.7 frontier0.959133/4062.1stextimagefile
44openai/gpt-5.5 frontier0.958733/4036.9stextimagefile
45openai/gpt-5.2-pro frontier0.957532/40180.7stextimagefile
46openai/gpt-6-astra frontier0.956735/4048.0stextimagefile
47deepseek/deepseek-v4.1-flash-10170.956334/4054.4stextimage
48openai/gpt-6.1-sol-pro frontier0.955435/4042.4stextimagefile
49anthropic/claude-opus-4.8 frontier0.954832/4076.6stextimagefile
50scaledown/extract0.954632/4076.0stextimagefile
51minimax/minimax-m3.1-flash:free0.953332/4077.3stextimagevideo
52moonshotai/kimi-k2.7-code0.953131/401144.9stextimage
53openai/gpt-5.2 frontier0.953131/40106.4stextimagefile
54deepseek/deepseek-v4-flash-07310.952732/4055.2stext
55deepseek/deepseek-v4-pro-08130.952133/40110.1stext
56xiaomi/mimo-v2.50.949633/4062.9stextimagevideoaudio
57nvidia/nemotron-3.5-super-vl:free0.949233/4045.0stextimagevideo
58typesafe/jev-router0.948334/4046.3stextimagefilevideoaudio
59openai/gpt-5.4-pro frontier0.947133/40256.4stextimagefile
60openai/gpt-6-astra-pro frontier0.946333/4069.4stextimagefile
61openai/gpt-6-sol frontier0.946030/4054.3stextimagefile
62openai/gpt-5.6-sol frontier0.945230/4039.2stextimagefile
63anthropic/claude-sonnet-5 frontier0.945233/40115.1stextimagefile
64openai/gpt-6.1-sol frontier0.945034/4048.4stextimagefile
65deepseek/deepseek-v4-flash0.945032/40150.6stext
66nvidia/nemotron-3.5-super-vl0.944931/40124.4stextimagevideo
67anthropic/claude-haiku-4.5 frontier0.944730/4058.8stextimagefile
68minimax/minimax-m2.10.942531/4060.0stext
69openai/gpt-5.6-sol-pro frontier0.942132/4053.2stextimagefile
70openai/gpt-5.6-luna frontier0.941231/4041.0stextimagefile
71openai/gpt-5.4-mini frontier0.940431/4072.3stextimagefile
72openai/gpt-5.1 frontier0.940330/40114.6stextimagefile
73openai/gpt-5.6-luna-pro frontier0.938530/4074.8stextimagefile
74meta/muse-glimmer-30b0.937933/4097.8stextimage
75anthropic/claude-haiku-5.5 frontier0.936432/4051.6stextimagefile
76openai/gpt-5-pro frontier0.932129/40354.8stextimagefile
77mistralai/mistral-large-4-00.932133/40188.9stextimage
78stepfun/step-3.5-flash0.927729/40165.8stext
79openai/gpt-6-sol-pro frontier0.926430/4076.3stextimagefile
80openai/gpt-5.4-nano frontier0.924727/40117.1stextimagefile
81openai/gpt-5.6-terra frontier0.924628/4026.9stextimagefile
82inclusionai/ling-3.0-flash0.924429/4048.6stext
83stepfun/step-3.7-flash0.924329/401536.5stextimagevideo
84minimax/minimax-m2.50.923728/40593.7stext
85openai/gpt-5 frontier0.920929/40131.1stextimagefile
86thinkingmachines/inkling0.920628/4047.2stextimageaudio
87openai/gpt-5.6-terra-pro frontier0.920027/4041.3stextimagefile
88minimax/minimax-m20.918729/4058.8stext
89upstage/solar-pro40.918330/4053.9stext
90reflection/beam-1.0-501b-a23b0.914230/4070.4stext
91tencent/hy4-preview0.911430/401548.2stext
92openai/gpt-6-luna frontier0.909528/4068.0stextimagefile
93openai/gpt-5-mini frontier0.909231/40120.1stextimagefile
94openai/gpt-chat-latest frontier0.898827/40119.8stextimagefile
95google/gemma-4-31b-it0.898326/40516.5stextimagevideo
96openai/gpt-6-luna-pro frontier0.878526/4078.7stextimagefile
97google/gemini-3-pro-image frontier0.872527/4065.8stextimageout:image
98google/gemma-4-26b-a4b-it0.858322/401001.6stextimagevideo
99openai/gpt-5-nano frontier0.789617/40106.5stextimagefile
100openai/o4-mini frontier0.772517/4053.2stextimagefile
101openai/o3 frontier0.720017/4043.1stextimagefile
102amazon/nova-premier-v1 frontier0.693916/40101.5stextimage
103amazon/nova-pro-v1 frontier0.61799/4047.7stextimage
03 · PARTIALS

14 partial runs, disclosed

These models stopped short of the full 40 tasks, mostly on slow responses that timed out. Ungraded tasks are excluded, never scored as zero, so these rows stay out of the ranking.

ModelTasks runMeanPerfectSpendStatusModalities
inclusionai/ling-3.1-flash23/400.989122partialtext
xiaomi/mimo-v2.6-pro38/400.989036partialtextimagevideoaudio
z-ai/glm-5v-turbo36/400.981933partialtextimagevideo
z-ai/glm-5.3-prime37/400.975933partialtext
z-ai/glm-5-turbo31/400.972628partialtext
aion-labs/aion-3.533/400.972229partialtext
xiaomi/mimo-v2.6-flash36/400.967632partialtextimagevideoaudio
poolside/laguna-s-2.139/400.948333partialtext
stealth/space-bunny-alpha38/400.947429partialtext
nex-agi/nex-n2-mini28/400.945823partialtext
aion-labs/aion-3.5-mini36/400.931729partialtext
nex-agi/nex-n2-pro14/400.928610partialtext
ibm-granite/granite-4.2-8b39/400.915825partialtext
liquid/lfm-2.5-24b-a2b34/400.884823partialtext
04 · CELL MATRIX

Every model, every task

The cell matrix shows one score per model per task: 103 ranked rows plus 14 partial rows across all 40 tasks, each cell linking to the graded answer.

Open the cell matrix

05 · DOWNLOADS

Take the data

Same run, two formats. Numbers match the boards above.

Spreadsheet

Results workbook

Leaderboard, full matrix, per-test difficulty, model info. Four sheets.

Download xlsx

Answers

Answer reviewer

Side-by-side: what each model wrote against the computed answer.

Open reviewer