Reported by APEX-Agents

APEX-Agents

Name: APEX-Agents results
Creator: APEX-Agents
License: https://www.mercor.com/apex/apex-agents-leaderboard/

Share of investment banking, consulting and corporate law tasks an agent completes in a simulated workplace with files and apps, graded against expert criteria (one attempt).

Last updated 1 Oct 2026

Results dated: 1 Oct 2026
Results: 39 configurations of 37 models
Unit: % of tasks
Licence: Published with permission (Mercor results via Epoch AI)

Tasks passed:

Top 15 of 39 results · % of tasks, higher is better. Choose a model to highlight it.Clear highlight

1 Claude Sonnet 5.5 (max reasoning)Anthropic 75.5%
2 Claude Opus 5.5 (max reasoning)Anthropic 73.5%
3 Claude Fable 5.1Anthropic 68.6%
4 Gemini 3.7 FlashGoogle 67.8%
5 Claude Opus 5 (max reasoning)Anthropic 65.8%
6 Grok 4.6xAI 65.3%
7 GPT-6 AstraOpenAI 64.7%
8 Gemini 3.8 FlashGoogle 64.3%
9 Claude Fable 5Anthropic 63.6%
10 GPT-6.1 Sol (max reasoning)OpenAI 60.0%
11 Claude Fable 5.1 (high reasoning)Anthropic 59.7%
12 GPT-5.6 Terra (max reasoning)OpenAI 58.2%
13 Muse Spark 1.3Meta 57.8%
14 GLM-5.3Z.ai 56.6%
15 Grok 4.5xAI 56.2%

Full results

APEX-Agents: tasks passed, % of tasks, higher is better
#	Model	Tasks passed % of tasks, higher is better	Price $ per million tokens, in / out
1	Claude Sonnet 5.5 (max reasoning)Anthropic	75.5%	$2 / $10
2	Claude Opus 5.5 (max reasoning)Anthropic	73.5%	$4 / $20
3	Claude Fable 5.1Anthropic	68.6%	$10 / $50
4	Gemini 3.7 FlashGoogle	67.8%	$1.50 / $7.50
5	Claude Opus 5 (max reasoning)Anthropic	65.8%	$5 / $25
6	Grok 4.6xAI	65.3%	$2 / $6
7	GPT-6 AstraOpenAI	64.7%	$10 / $50
8	Gemini 3.8 FlashGoogle	64.3%	$1.50 / $7.50
9	Claude Fable 5Anthropic	63.6%	$10 / $50
10	GPT-6.1 Sol (max reasoning)OpenAI	60.0%	$2 / $10
11	Claude Fable 5.1 (high reasoning)Anthropic	59.7%	$10 / $50
12	GPT-5.6 Terra (max reasoning)OpenAI	58.2%	$2 / $12
13	Muse Spark 1.3Meta	57.8%	$1.25 / $4.25
14	GLM-5.3Z.ai	56.6%	$1.40 / $4.40
15	Grok 4.5xAI	56.2%	$2 / $6
16	GPT-5.5OpenAI	55.1%	$5 / $30
17	Claude Sonnet 5Anthropic	54.5%	$2 / $10
18	GPT-6 Sol (max reasoning)OpenAI	54.3%	$2 / $10
19	GLM-5.3-FlashZ.ai	52.8%	$0.15 / $0.50
20	GPT-5.4OpenAI	52.4%	$2.50 / $15
21	GPT-5.6 Sol Pro (max reasoning)OpenAI	51.4%	$4 / $20
22	Kimi K3Moonshot AI	50.6%	$3 / $15
23	Claude Opus 4.7 (max reasoning)Anthropic	49.2%	$5 / $25
24	Claude Opus 4.8 (max reasoning)Anthropic	48.9%	$5 / $25
25	DeepSeek-V4-Pro (0813)DeepSeek	47.3%	$1.32 / $3.96
26	Gemini 3.6 FlashGoogle	46.9%	$1.50 / $7.50
27	Claude Opus 4.6 (max reasoning)Anthropic	46.3%	$5 / $25
28	Claude Sonnet 5.5 (medium reasoning)Anthropic	44.6%	$2 / $10
29	GPT-6 Luna (max reasoning)OpenAI	44.3%	$0.10 / $0.50
30	Claude Sonnet 4.6 (high reasoning)Anthropic	43.0%	$3 / $15
31	GLM-5.1Z.ai	40.9%	$1.38 / $4.40
32	MiniMax-M3MiniMax	37.7%	$0.30 / $1.20
33	Kimi K2.7 CodeMoonshot AI	37.6%	$0.95 / $4
34	Muse Spark 1.2Meta	36.4%	$1.25 / $4.25
35	Gemini 3.1 Pro PreviewGoogle	35.3%	$2 / $12
36	InklingThinking Machines	33.8%	$0.95 / $4.05
37	Gemini 3.5 FlashGoogle	27.5%	$1.50 / $9
38	Qwen3.5-397B-A17BAlibaba	24.9%	$0.55 / $3.50
39	gpt-oss-120bOpenAI	4.4%	$0.15 / $0.60

Swipe the table sideways for more columns.

Ranks follow the score as shown, so equal numbers share a rank. Results as published by APEX-Agents; we do not re-run them.

Progress over time

Tasks passed, plotted at each model's release date, higher is better. Each model counts once, at its best setting. Scores are today's, not launch day's; 1 model without a known release date is left off. Hover or tap a mark for its model.

Best score so far

Every model at its release date. The line moves each time a new release beats every earlier one, and each new best is named where there is room.

Best score by lab

Each lab's best so far, stepping whenever one of its releases beats its own earlier models. The 6 labs with the best scores today; 5 others not drawn.

AnthropicGooglexAIOpenAIMetaZ.ai

What it measures

Share of investment banking, consulting and corporate law tasks an agent completes in a simulated workplace with files and apps, graded against expert criteria (one attempt).

What it does not measure

Not your firm's documents or tools; graded by rubric, not by a client.

Source

APEX-Agents · Methodology

Mercor; collected by Epoch AI. Licence: Published with permission (Mercor results via Epoch AI).