APEX-Agents leaderboard
As of November 8, 2026, Kimi K3.1 Pro (Moonshot) leads APEX-Agents at 69.4%, ahead of Grok 4.6 at 57.5% and Grok 4.5 at 47.1%. 6 systems ranked. Every score here is published by the model's own lab; each row links to the page it comes from.
What is APEX-Agents? APEX-Agents (AI Productivity Index for Agents) tests whether an agent can do long, cross-application work set by investment banking analysts, management consultants and corporate lawyers. Made by Mercor, January 2026.
Top models on APEX-Agents, November 2026
| # | Model | Score | Price / 1M | Context | Source |
|---|---|---|---|---|---|
| 1 | Kimi K3.1 Pro Moonshot (Kimi) | max effort | $2 / $10 | 1M | Vendor-reported Moonshot AI |
| 2 | Grok 4.6 SpaceXSI | high effort | $2 / $6 | 500K | Vendor-reported SpaceXSI (xAI) |
| 3 | Grok 4.5 SpaceXSI | high effort | $2 / $6 | 500K | Vendor-reported SpaceXSI (xAI) |
| 4 | Kimi K3 Moonshot (Kimi) | max effort | $3 / $15 | 1M | Vendor-reported Moonshot AI |
| 5 | Gemini 3.1 Pro Preview Preview Google · listed as “Gemini 3.1 Pro” | thinking high | $2 / $12 | 1M | Vendor-reported Google DeepMind |
| 6 | Gemini 3 Pro Preview Retired Google · listed as “Gemini 3 Pro” | thinking high | $2 / $12 | 1M | Vendor-reported Google DeepMind |
APEX-Agents: score against price
Can you trust APEX-Agents scores?
No independent leaderboard publishes APEX-Agents results for these models yet, so every number on this page is the lab's own report of its own model. Labs choose the harness, the effort level and the subset they report, and they rarely publish failures. Read the setting under each score before comparing two rows.
- Mercor builds the benchmark, runs the evaluation and hosts the leaderboard.
- Grading is done by a model judge.
- Mercor publishes Pass@1 and a mean rubric score; check which one a lab quotes.
How APEX-Agents works
The task
The agent works inside a realistic workspace with files and tools and has to deliver a finished work product, such as an analysis or a memo, that spans several applications.
The score
A model judge grades each output against expert pass-or-fail rubrics. Pass@1 is the share of tasks fully passed on one attempt.
Test set
480 tasks (open-source release), as published by the authors.
Made by
Mercor, first published January 2026.
Official sources
- www.mercor.com/apex/apex-agents-leaderboard
- Paper: arxiv.org/abs/2601.14242
- Definition read on 2026-11-08
Sources and freshness
6 published scores from 4 sources. Each was read on its source page; the oldest reading on this page is from November 8, 2026. Prices beside each score come from our model price index and move when a vendor reprices.
- Vendor-reported Google DeepMind
- Vendor-reported SpaceXSI (xAI)
- Vendor-reported Moonshot AI
- Vendor-reported Moonshot AI
APEX-Agents FAQ
Answers derived from the 6 published scores on this page and the current price index.