APEX Agents benchmark
26 AI models ranked on APEX Agents. Each row shows the model’s score next to its cheapest API price per 1M tokens where it has one, so you can weigh quality against cost.
Top model today
Gemini 3.5 Flash (Google) leads at 47%, from $1.50 per 1M input.
Data via arXiv:2601.14242 →Models ranked
26
on this benchmark
Top score
47%
current leader
Type
Single eval
benchmark
Updated
15 July 2026
last source fetch
Cost vs APEX Agents
21 priced modelsEach labelled name links to that model’s page. Curated to the cost-efficiency frontier, top scorers and cheapest so labels stay readable; the full field is in the leaderboard table below (and via “select all”). Reasoning models (extended “thinking” before answering) carry a ring around the dot.
Models ranked on APEX Agents
highest score first · 26 models| # | ||||||
|---|---|---|---|---|---|---|
| 1 | Gemini 3.5 Flash |  Google | 47% | $1.50 | $9.00 | |
| 2 | GPT-5.5 |  OpenAI | 38% | $5.00 | $30.00 | |
| 3 | GLM 5.2 |  Zhipu | 34% | $1.40 | $4.40 | |
| 4 | GPT-5.4 | OpenAI | 33% | $2.50 | $15.00 | |
| 5 | Claude Opus 4.6 |  Anthropic | 33%reasoning: adaptive | $5.00 | $25.00 | |
| 6 | Gemini 3.1 Pro Preview | Google | 32% | $2.00 | $12.00 | |
| 7 | Kimi K2.6 |  Moonshot | 28% | n/a | n/a | |
| 8 | GPT-5.4 Mini | OpenAI | 28% | $0.75 | $4.50 | |
| 9 | Claude Sonnet 4.6 | Anthropic | 28%reasoning: adaptive | $3.00 | $15.00 | |
| 10 | Gemini 3 Flash | Google | 28%reasoning: true | n/a | n/a | |
| 11 | GPT-5.4 Nano | OpenAI | 25% | $0.20 | $1.25 | |
| 12 | DeepSeek V4 Pro |  DeepSeek | 24% | $0.43 | $0.87 | |
| 13 | Qwen3.7 Plus |  Alibaba | 22% | n/a | n/a | |
| 14 | Grok 4.3 |  xAI | 17% | $1.25 | $2.50 | |
| 15 | Qwen3.5 397B A17B | Alibaba | 15% | $0.60 | $3.60 | |
| 16 | Step 3.7 Flash | Stepfun | 15% | n/a | n/a | |
| 17 | DeepSeek V3.2 | DeepSeek | 15%reasoning: true | $0.27 | $0.40 | |
| 18 | GLM 5 | Zhipu | 14% | $0.80 | $2.56 | |
| 19 | Grok 4.20 0309 | Spacexai | 14% | n/a | n/a | |
| 20 | Gemini 3.1 Flash Lite Preview | Google | 12% | $0.25 | $1.50 | |
| 21 | Kimi K2.5 | Moonshot | 12% | $0.50 | $2.80 | |
| 22 | MiniMax M2.7 |  MiniMax | 11% | $0.26 | $0.55 | |
| 23 | gpt-oss-120b | OpenAI | 3% | $0.05 | $0.25 | |
| 24 | MiMo-V2.5-Pro |  Xiaomi | 2% | $1.00 | $3.00 | |
| 25 | Nemotron 3 120B A12b |  NVIDIA | 2% | $0.50 | $1.50 | |
| 26 | gpt-oss-20b | OpenAI | 1% | $0.01 | $0.07 |
Frequently asked questions
What is the APEX Agents benchmark?
APEX (AI Productivity Index for Agents) is a benchmark of 480 long-horizon, cross-application tasks written by investment-banking analysts, management consultants and corporate lawyers, testing whether agents can autonomously complete realistic professional workflows using files and tools. Graded against task-specific rubrics.
Which AI model scores highest on APEX Agents?
Gemini 3.5 Flash (Google) leads with 47%, from $1.50 per 1M input tokens.
How many models are ranked on APEX Agents?
26 models carry a APEX Agents score. Scores come from Artificial Analysis; where a model has a live API price it is verified against LiteLLM and OpenRouter, and models with no current price are listed without one.
Other benchmarks
compare the same models on a different evalEvery weekday
AI moves fast. Here's your debrief.
News, analysis, tools, and more.
For people who build with AI