Skip to content
Block4Roots

Block4Roots · AI leaderboards

Which AI is winning right now?

Simple rankings from public data: what people use most, what is best at each job, and what gives the most for your money.

Auto-updated · last refresh 18 Oct, 09:47 pm

Most used AI this week (so far)

Ranked by each model's share of all AI text sent that week. Arrows compare its position with last week.

  1. #1

    DeepSeek V4.1 Flash

    20% –
  2. #2

    GLM-5.3-Flash

    7.4% –
  3. #3

    Step 5 Preview

    7.1% ↑
  4. #4

    MiMo-V2.6-Flash

    6.1% ↓
  5. #5

    Hy4 preview

    4.6% –
  6. #6

    GPT-6 Luna

    3.8% –
  7. #7

    Claude Opus 5.5

    3.8% ↑
  8. #8

    GLM-5.3

    3.2% ↑

Source: OpenRouter (openrouter.ai/rankings), as of 18 Oct 2026

Best for…

The top scorer in each benchmark (test). Scores are like an intelligence score, out of 100.

Coding

Claude Fable 5.1

81.6 score

Score on coding tests, out of 100.

Overall smarts

Claude Opus 5.5

57.6 score

Average score across many reasoning and knowledge tests.

Agents and tasks

Claude Fable 5.1

57.9 score

How well it finishes multi-step jobs on its own.

Websites

Gemini 3 Pro Preview

64% win rate

Win rate = how often people picked its design over another AI's.

Hard science questions

Gemini 4 Argon

96%

Share of very hard science questions it got right.

Customer-support agent tasks

Claude Fable 5

80%

Share of airline support chats it handled correctly.

Web search

Claude Opus 5

89%

How well it finds correct answers by searching the web.

Best for design work

Win rate = how often people picked its design over another AI's.

Websites

Gemini 3 Pro Preview

64% wins

App screens

Gemini 3 Pro Preview

69% wins

Logos

Riverflow 2.5 Pro

64% wins

Images

Riverflow 2.5 Pro

68% wins

Games

Claude Sonnet 5.5

73% wins

Charts

Gemini 2.5 Pro

68% wins

Source: Artificial Analysis, Design Arena, OpenRouter benchmarks via OpenRouter (openrouter.ai/rankings), as of 17 Oct 2026

Best value

Smarts score divided by API price per 1M tokens. Higher means more brains per rupee. Price mixes 3 parts input to 1 part output.

  1. #1GLM-5.3-FlashBest valuescore 42 · $0.08 per 1M tokens
  2. #2Claude Haiku 5.5score 43 · $0.22 per 1M tokens
  3. #3Ling 3.1 Flashscore 41 · $0.32 per 1M tokens
  4. #4MiMo-V2.6-Proscore 46 · $0.54 per 1M tokens
  5. #5Step 5 Previewscore 44 · $1.4 per 1M tokens

Source: Artificial Analysis via OpenRouter (openrouter.ai/rankings), as of 17 Oct 2026

What people use AI for

Share of all AI use by kind of job. Top model = the one used most for that job.

  • General59%
  • Data18%
  • Agent14%
  • Code10%
  • Classification 32%Top model: Jev 1.13
  • Data Extraction 13%Top model: GLM 5.3 Flash
  • Workflow Execution 8.4%Top model: DeepSeek V4.1 Flash
  • Roleplay & Fiction 7.4%Top model: DeepSeek V4.1 Flash
  • Content Writing 5.4%Top model: GLM 5.3 Flash

Source: OpenRouter (openrouter.ai/rankings), as of 17 Oct 2026

Trending AI apps

Apps growing fastest on OpenRouter right now. Underlined names have a Block4Roots page.

  1. #1Claude Code
  2. #2Kilo Code
  3. #3Cline
  4. #4Hermes Agent
  5. #5Codex
  6. #6pi
  7. #7omp
  8. #8Freebuff
  9. #9DeepSeek Harness
  10. #10Command Code

Source: OpenRouter (openrouter.ai/rankings), as of 18 Oct 2026

AI video makers: specs

Longest videos first. Max length = longest clip in one go; quality = sharpest picture.

Wan 3.0 Prime

Max length
30 s
Best quality
1080p
Sound
Yes

Wan 3.0

Max length
30 s
Best quality
1080p
Sound
Yes

Seedance 2.5

Max length
30 s
Best quality
720p
Sound
Yes

FLUX 3 Video

Max length
20 s
Best quality
4K
Sound
Yes

Grok Imagine Video 1.5 Lite

Max length
15 s
Best quality
1080p
Sound
—

HeyGen Video

Max length
15 s
Best quality
768p
Sound
No

H3 Max

Max length
15 s
Best quality
768p
Sound
No

Seedance 2.0 Mini

Max length
15 s
Best quality
720p
Sound
Yes

H3

Max length
15 s
Best quality
2K
Sound
Yes

Source: OpenRouter (openrouter.ai/rankings), as of 18 Oct 2026