Agentic

🏆 Leaderboard

As of August 22, 2026, Kimi K3 is #1 for agent and computer-use work at 91.2%. Ranked by the BrowseComp score Best AI agent ranked by OSWorld, Claude computer use, tau-bench, and tool-use evals. Compare computer-use scores next to API price.

Updated August 22, 2026282 models33 providers


Kimi K3OSS Moonshot AI · Open Source	91.2%	37.6%	—	—	84.8%	76.5%	84.2%	—	—	—	—	—	—	—	$18.00
Claude Opus 5 Anthropic · Proprietary	90.8%	43.5%	—	—	70.6%	80.6%	85.8%	83.4%	—	60.4%	—	—	70.6%	—	$30.00
GPT-5.6 Sol OpenAI · Proprietary	90.4%	39.9%	—	—	—	58%	—	—	—	54.1%	—	—	62.6%	52.7%	$35.00
GPT-5.5 Pro OpenAI · Proprietary	90.1%	—	—	—	—	—	—	—	—	—	—	—	—	—	$540.00
Claude Mythos 5 Anthropic · Proprietary	88%	—	—	—	85%	61.7%	—	85%	—	—	—	—	—	—	$60.00
GPT-5.6 Terra OpenAI · Proprietary	87.5%	—	—	—	—	53.1%	—	—	—	60.6%	—	—	50.2%	50.4%	$14.00
Claude Mythos Preview Anthropic · Proprietary	86.9%	—	—	—	—	—	—	79.6%	—	—	—	—	—	—	$60.00
Kimi K2.6OSS Moonshot AI · Open Source	86.3%	27.9%	—	—	—	50%	—	73.1%	—	—	—	—	4.6%	—	$4.93
Gemini 3.1 Pro Google · Proprietary	85.9%	33.5%	—	—	—	—	69.2%	78.3%	—	—	—	—	—	—	$17.50
Seed 2.1 Turbo ByteDance · Proprietary	84.9%	29.2%	—	—	76.4%	49.1%	80.3%	—	—	—	—	—	—	—	$3.00
Claude Sonnet 5 Anthropic · Proprietary	84.7%	32.5%	—	—	81.2%	74.7%	—	81.2%	—	46.5%	—	—	—	—	$12.00
GPT-5.5 OpenAI · Proprietary	84.4%	38.5%	—	—	78.7%	55.6%	75.3%	78.7%	98%	62.2%	—	54.0%	13%	—	$35.00
Claude Opus 4.8 Anthropic · Proprietary	84.3%	42.5%	—	—	—	59.9%	82.2%	83.4%	—	59.2%	87.9%	50.2%	20.6%	—	$30.00
Hy3OSS Tencent · Open Source	84.2%	25.6%	—	—	—	48.5%	79.1%	—	—	55.3%	—	—	—	—	$0.66
Claude Opus 4.6 Anthropic · Proprietary	84%	32.4%	91.9%	—	72.7%	—	62.7%	—	99.3%	—	—	55.3%	—	—	$30.00
MiniMax M3OSS MiniMax · Open Source	83.5%	27.7%	—	—	—	—	74.2%	70.1%	—	51.5%	—	—	4.6%	—	$1.50
DeepSeek-V4-Pro-MaxOSS DeepSeek · Open Source	83.4%	—	—	—	—	51.8%	73.6%	—	—	—	—	—	—	—	$5.22
GPT-5.6 Luna OpenAI · Proprietary	83.3%	—	—	—	—	53.4%	—	—	—	60.5%	—	—	45.6%	50.3%	$1.40
GPT-5.4 OpenAI · Proprietary	82.7%	36%	—	—	64.9%	54.6%	67.2%	75%	71%	51.7%	—	35.1%	—	—	$17.50
GLM-5.1OSS Z AI · Open Source	79.3%	—	—	—	—	40.7%	71.8%	—	—	—	—	—	—	—	$5.80
Claude Opus 4.7 Anthropic · Proprietary	79.3%	33.9%	—	—	—	—	77.3%	78%	—	—	—	—	18.2%	—	$30.00
GPT-5.2 Pro OpenAI · Proprietary	77.9%	—	—	—	47.8%	—	77.8%	—	—	—	—	—	—	—	$189.00
Inkling-SmallOSS Thinking Machines · Open Source · via Thinking Machines Lab	77.4%	—	—	—	—	54.4%	79.6%	—	—	33.6%	—	—	—	—	$1.50
Seed 2.0 Pro ByteDance · Proprietary	77.3%	—	—	—	—	—	—	—	—	—	—	—	—	—	$3.50
MiniMax M2.5OSS MiniMax · Open Source	76.3%	6.2%	—	—	—	—	—	—	—	—	—	—	—	—	$1.50
GLM-5OSS Z AI · Open Source	75.9%	17.2%	—	—	—	—	67.8%	—	—	—	—	—	—	—	$4.20
Kimi K2.5OSS Moonshot AI · Open Source	74.9%	14.4%	—	—	63.3%	—	—	63.3%	—	—	—	—	—	—	$3.68
Claude Sonnet 4.6 Anthropic · Proprietary	74.7%	23.7%	91.7%	—	78.5%	—	61.3%	72.1%	97.9%	49.0%	—	54.9%	9.3%	—	$18.00
DeepSeek-V4-Flash-MaxOSS DeepSeek · Open Source	73.2%	—	—	—	—	47.8%	69%	—	—	—	—	—	—	—	$0.42
Step-3.5-FlashOSS StepFun · Open Source	69%	—	—	—	—	—	—	—	—	—	—	—	—	—	$0.50
Qwen3.5-397B-A17BOSS Qwen · Open Source	69%	—	—	—	—	38.3%	—	—	—	—	—	—	—	—	$4.20
GPT-5.2 OpenAI · Proprietary	65.8%	34.4%	82%	55.9%	39.2%	46.3%	60.6%	—	98.7%	—	86.3%	41.1%	—	—	$15.75
Qwen3.5-122B-A10BOSS Qwen · Open Source	63.8%	—	—	—	—	—	—	58%	—	—	70.4%	—	—	—	$3.60
MiniMax M2.1OSS MiniMax · Open Source	62%	—	—	—	—	43.5%	—	—	87%	—	—	—	—	—	$1.50
Qwen3.5-35B-A3BOSS Qwen · Open Source	61%	—	—	—	—	—	—	54.5%	—	—	68.6%	—	—	—	$2.25
Qwen3.5-27BOSS Qwen · Open Source	61%	—	—	—	—	—	—	56.2%	—	—	70.3%	—	—	—	$2.70
Kimi K2-Thinking-0905OSS Moonshot AI · Open Source	60.2%	—	—	—	—	—	—	—	—	—	—	—	—	—	$2.47
MiMo-V2-FlashOSS Xiaomi · Open Source	58.3%	—	—	—	—	—	—	—	—	—	—	—	—	—	$0.40
LongCat-Flash-Thinking-2601OSS Meituan · Open Source	56.6%	—	88.6%	—	—	—	—	—	99.3%	—	—	—	—	—	$1.50
GPT-5 OpenAI · Proprietary	54.9%	18.3%	81.1%	—	—	—	—	53.6%	96.7%	—	—	55.1%	—	—	$11.25
DeepSeek-V4-Flash-0423OSS DeepSeek · Open Source	53.5%	—	—	—	—	43.5%	67.4%	—	—	—	—	—	—	—	$0.30
GLM-4.7OSS Z AI · Open Source	52%	8.7%	—	—	—	—	—	—	—	—	—	—	—	—	$2.80
o4-mini OpenAI · Proprietary	51.5%	—	71.8%	53.2%	—	—	—	—	—	—	—	—	—	—	$5.50
DeepSeek-V3.2OSS DeepSeek · Open Source · via OpenRouter	51.4%	7%	—	—	—	35.2%	—	—	—	—	—	—	—	—	$0.57
o3 OpenAI · Proprietary	49.7%	17.2%	80.2%	63.0%	23%	—	—	23%	58.2%	—	—	46.6%	—	—	$10.00
Solar Pro 4 Upstage · Proprietary	49.2%	—	—	—	—	—	—	—	—	—	—	—	—	—	$1.50
Mistral Medium 3.5OSS Mistral · Open Source · via Mistral AI	48.6%	—	—	—	—	—	—	—	—	—	—	—	—	—	$9.00
GLM-4.6OSS Z AI · Open Source	45.1%	4%	—	72.4%	—	—	—	—	—	—	—	—	—	—	$2.80
Grok 4 Fast xAI · Proprietary	44.9%	—	—	—	—	—	—	—	—	—	—	—	—	—	$0.70
Nemotron 3 Ultra (550B A55B)OSS NVIDIA · Open Source · via OpenRouter	44.4%	—	—	—	—	—	—	—	—	—	—	—	—	—	$2.70

Showing 1–50 of 282 models

Next step

You found the model. Now ship the product.

Auth, billing, and the API layer are already decided. Fork 8 finished AI apps, or have us build the first version with you.

Build with the APIs Have us build it with you

The index

All Large Language Models

282 models across 33 providers. Search or jump to a lab — every model page stays linked here.

Cursor

2 models

Composer 2 Composer 2 Fast

xAI

15 models

Grok 4.6 Grok 4.5 Grok 4.3 Grok-4.1 Grok-4.1 Fast Non-Reasoning Grok-4.1 Fast Reasoning Grok-4.1 Thinking Grok 4 Fast Grok Code Fast 1 Grok-4 Fast Non-Reasoning Grok-4 Fast Reasoning Grok-4 Grok-3 Grok-3 Mini Grok-2

DeepSeek

16 models

DeepSeek-V4-Pro-0813 DeepSeek-V4-Flash-0731 DeepSeek-V4-Flash-0423 DeepSeek-V4-Flash-Max DeepSeek-V4-Pro-Max DeepSeek-V3.2 DeepSeek-V3.2 (Non-thinking)DeepSeek-V3.2-Exp DeepSeek-R1-0528 DeepSeek-V3 0324 DeepSeek R1 Distill Llama 70B DeepSeek R1 Distill Qwen 32B DeepSeek-R1 DeepSeek-V3.1 DeepSeek-V3 DeepSeek-V2.5

MiniMax

6 models

MiniMax M3 MiniMax M2.7 MiniMax M2.5 MiniMax M2.1 MiniMax M2 MiniMax M1 80K

AI21 Labs

2 models

Jamba 1.5 Large Jamba 1.5 Mini

Amazon

5 models

Nova 2 Lite Nova 2 Sonic Nova Lite Nova Micro Nova Pro

Baidu

1 models

ERNIE 4.5

ByteDance

2 models

Seed 2.1 Turbo Seed 2.0 Pro

Cohere

3 models

North Mini Code 1.0 Command A+Command R+

IBM

1 models

Granite 3.3 8B Instruct

Inception

1 models

Mercury 2

inclusionAI

1 models

Ling-3.0-flash

LG AI Research

1 models

K-EXAONE-236B-A23B

Meituan

4 models

LongCat-Flash-Lite LongCat-Flash-Thinking-2601 LongCat-Flash-Thinking LongCat-Flash-Chat

Microsoft

5 models

MAI-Code-1.1-Flash Phi 4 Mini Phi-4-multimodal-instruct Phi 4 Phi-3.5-mini-instruct

Moonshot AI

9 models

Kimi K3 Kimi K2.7 Code Kimi K2.7 Code HighSpeed Kimi K2.6 Kimi K2.5 Kimi K2 0905 Kimi K2-Instruct-0905 Kimi K2-Thinking-0905 Kimi K2 Instruct

Nous Research

1 models

Hermes 3 70B

NVIDIA

5 models

Nemotron 3.5 Lightning (30B A3B)Nemotron 3 Ultra (550B A55B)Nemotron 3 Super (120B A12B)Nemotron 3 Nano (30B A3B)Nemotron Nano 9B v2

Poolside

2 models

Laguna S 2.1 Laguna XS 2.1

Sakana AI

1 models

Sakana Namazu

StepFun

1 models

Step-3.5-Flash

Tencent

1 models

Hy3

Thinking Machines

1 models

Inkling-Small

Unisound

1 models

Upstage

1 models

Solar Pro 4

Xiaomi

3 models

MiMo-V2.5-Pro MiMo-V2.5 MiMo-V2-Flash

What is the best AI agent?

As of August 22, 2026, Kimi K3 by Moonshot AI is #1 for agent and computer-use work at 91.2%. Ranked by the BrowseComp score This board also tracks BrowseComp, APEX-Agents, tau-bench Retail, BFCL. Next on the same board: Claude Opus 5 and GPT-5.6 Sol. Related leaders: Grok 4.6 on APEX-Agents at 57.5%; Claude Opus 4.6 on tau-bench Retail at 91.9%. This agentic leaderboard ranks models by BrowseComp. Scores come from public evals. Prices are the live API rates in the table above.

Sources: BrowseComp (OpenAI) (https://openai.com/index/browsecomp/); OpenAI API pricing (https://developers.openai.com/api/docs/pricing); Anthropic Claude API pricing (https://platform.claude.com/docs/en/about-claude/pricing); Gemini API pricing (https://ai.google.dev/gemini-api/docs/pricing)

Top 8 for agentic. Ranked by the BrowseComp score Input and output are dollars per million tokens.
Rank	Model	BrowseComp	Input /M	Output /M
1	Kimi K3	91.2%	$3.00	$15.00
2	Claude Opus 5	90.8%	$5.00	$25.00
3	GPT-5.6 Sol	90.4%	$5.00	$30.00
4	GPT-5.5 Pro	90.1%	$60.00	$480.00
5	Claude Mythos 5	88%	$10.00	$50.00
6	GPT-5.6 Terra	87.5%	$2.00	$12.00
7	Claude Mythos Preview	86.9%	$10.00	$50.00
8	Kimi K2.6	86.3%	$0.96	$3.97

What is Claude computer use?

Claude computer use is the Anthropic name for letting Claude drive a computer: screenshot in, mouse and keyboard out. People search it more than “best LLM for agents.” OSWorld is the public exam for that loop. A high chat score does not mean the model can finish a desktop task.

OSWorld vs tau-bench

OSWorld is GUI computer use. tau-bench is multi-turn tool calls in a text workflow. BrowseComp is web research. Sort the column that matches the loop you ship. Mixing them into one “best agent” trophy is how teams buy the wrong SKU.

Related LLM leaderboards

LLM Leaderboard Coding Math Reasoning Writing Science Finance Long Context Vision Open Source

Individual benchmark pages

Rank one eval at a time. All LLM benchmarks.

HLE SWE Bench Humanity's Last Exam (No Tools)Humanity's Last Exam (With Tools)Terminal-Bench 2.0 Terminal-Bench Terminal-Bench 2.1 Terminal-Bench 2 Terminal-Bench 3.0 ARC-AGI-1 Verified

Agentic FAQ

Who ranks #1 on the Agentic leaderboard?

As of August 22, 2026, Kimi K3 by Moonshot AI ranks #1 on BrowseComp at 91.2%. API pricing is $3.00/M input and $15.00/M output.

What is the best AI agent?

What are the top models on BrowseComp?

The current BrowseComp ranking as of August 22, 2026 is 1. Kimi K3 at 91.2%; 2. Claude Opus 5 at 90.8%; 3. GPT-5.6 Sol at 90.4%.

Which agentic model is the cheapest?

Nemotron 3.5 Lightning (30B A3B) is the cheapest scored model on this agentic leaderboard at $0.05/M input and $0.20/M output ($0.25 blended). Kimi K3 still leads BrowseComp at 91.2%.

Should I always pick the #1 BrowseComp model?

Not automatically. Kimi K3 leads BrowseComp, but a cheaper scored model can be the better production choice if the quality gap is small. Use the table to weigh BrowseComp against input/output price, context window, and related evals.

How often is the Agentic leaderboard updated?

Scores and API prices on this page are refreshed from published evals and provider rates. The snapshot is labeled August 22, 2026. Treat it as a current index, not a one-off blog post.

What is the best AI agent model?

OSWorld ranks desktop computer-use, tau-bench ranks multi-turn tool conversations, and BrowseComp ranks web research. Sort by the agent loop you ship.

What is OSWorld?

OSWorld is a computer-use benchmark. Models have to click, type, and finish real desktop tasks instead of answering a chat prompt.

What is tau-bench?

tau-bench and tau2-bench score agents on retail, airline, and telecom workflows that need reliable tool calls over many turns.

What is Claude computer use?

Claude computer use is Anthropic's desktop-control loop: the model sees the screen, then clicks and types. OSWorld is the usual public score for that job. Computer use is not the same as a chat model with function calling.

What is computer use in AI?

Computer use means the agent operates a real desktop or browser: mouse, keyboard, apps. OSWorld and OSWorld-Verified score that. Tool-calling benches like tau-bench score API tools in a conversation, not a GUI.

Is the best AI agent the same as the best chatbot?

No. Chat quality and agent success diverge. A model can win Arena Elo and fail OSWorld. Rank this board for agents, the overall board for chat, and coding for SWE-bench.

You found the model. Now ship the product.

All Large Language Models

OpenAI

Anthropic

Google

Cursor

xAI

DeepSeek

MiniMax

Qwen

AI21 Labs

Amazon

Baidu

ByteDance

Cohere

IBM

Inception

inclusionAI

LG AI Research

Meituan

Meta

Microsoft

Mistral

Moonshot AI

Nous Research

NVIDIA

Poolside

Sakana AI

StepFun

Tencent

Thinking Machines

Unisound

Upstage

Xiaomi

Z AI

What is the best AI agent?

What is Claude computer use?

OSWorld vs tau-bench

Related LLM leaderboards

Individual benchmark pages

Agentic FAQ

Who ranks #1 on the Agentic leaderboard?

What is the best AI agent?

What are the top models on BrowseComp?

Which agentic model is the cheapest?

Should I always pick the #1 BrowseComp model?

How often is the Agentic leaderboard updated?

What is the best AI agent model?

What is OSWorld?

What is tau-bench?

What is Claude computer use?

What is computer use in AI?

Is the best AI agent the same as the best chatbot?