AI model leaderboard
Top today: Gemini 4 Carbon. The smartest chat models ranked by blind human votes, with what each costs developers per million tokens.
Smartest first.
Arena score from LMArena (text). Prices from OpenRouter, the cheapest provider listed. Looking for an app to use? See the best chat assistants.
| # | Model | Arena score | Input $/1M | Output $/1M |
|---|---|---|---|---|
| 01 | Gemini 4 CarbonGoogle | 1540 | n/a | n/a |
| 02 | Gemini 4 ArgonGoogle | 1529 | $2 | $10 |
| 03 | Claude Opus 5.5Anthropic | 1516 | $4 | $20 |
| 04 | Claude Fable 5.1Anthropic | 1514 | $10 | $50 |
| 05 | Claude Opus 5Anthropic | 1508 | $5 | $25 |
| 06 | Claude Opus 4.6Anthropic | 1502 | $5 | $25 |
| 07 | Muse WonderMeta | 1499 | $4 | $20 |
| 08 | Gemini 3.8 FlashGoogle | 1498 | $0.82 | $4.13 |
| 09 | Claude Fable 5Anthropic | 1492 | $10 | $50 |
| 10 | MiMo-V2.6-ProXiaomi | 1491 | $0.43 | $0.87 |
| 11 | Muse Spark 1.3Meta | 1491 | $1.25 | $4.25 |
| 12 | Claude Opus 4.7Anthropic | 1490 | $5 | $25 |
| 13 | Gemini 3.7 FlashGoogle | 1489 | $0.75 | $3.75 |
| 14 | Qwen3.8-MaxAlibaba | 1483 | $2 | $6 |
| 15 | Muse Spark 1.2Meta | 1483 | $1.25 | $4.25 |
| 16 | Gemini 3.5 FlashGoogle | 1482 | $1.50 | $9 |
| 17 | Gemini 3.1 Pro PreviewGoogle | 1480 | $2 | $12 |
| 18 | Gemini 3.6 FlashGoogle | 1480 | $0.75 | $3.75 |
| 19 | Gemini 3 ProGoogle | 1479 | n/a | n/a |
| 20 | Muse Spark 1.1Meta | 1478 | $1.25 | $4.25 |
| 21 | Kimi K3Moonshot AI | 1476 | $0.19 | $16 |
| 22 | Qwen3.7-Max-PreviewAlibaba | 1475 | n/a | n/a |
| 23 | Muse SparkMeta | 1474 | $1.25 | $4.25 |
| 24 | GLM-5.3Z.ai | 1474 | $0.07 | $1.40 |
| 25 | GLM-5.3-FlashZ.ai | 1473 | $0.15 | $0.50 |
Sources: LMArena text leaderboard (CC-BY-4.0), OpenRouter prices. Updated by our tracker every day.
Who leads in images, video and voice.
Top five on each public leaderboard we track. Scores are votes from people comparing results blind, except transcription, which measures mistakes.
Making images
Our picks- 1GPT-Image-2.5 Sunburst1427
- 2GPT-Image-2.6 Sunburst1409
- 3GPT-Image-2.5 Flare1399
- 4GPT-Image-2.6 Flare1384
- 5GPT-Image-21382
LMArena text to image · 2026-11-04
Editing photos
Our picks- 1GPT-Image-2.5 Sunburst1517
- 2GPT-Image-2.5 Flare1482
- 3GPT-Image-2.6 Sunburst1476
- 4MAI-Image-2.7-Preview1469
- 5GPT-Image-2.6 Flare1462
LMArena image edit · 2026-11-04
Video from text
Our picks- 1gemini-omni-1.1-flash1516
- 2gemini-omni-flash1513
- 3flux-3-video1493
- 4grok-imagine-video-1.5-agent1492
- 5dreamina-seedance-2.0-720p1479
LMArena text to video · 2026-09-22
Video from a photo
Our picks- 1minimax-h31495
- 2gemini-omni-1.1-flash1488
- 3wan3.01480
- 4dreamina-seedance-2.5-720p1477
- 5dreamina-seedance-2.0-720p1475
LMArena image to video · 2026-09-22
Voices (text to speech)
Our picks- 1Inworld TTS MAX1545
- 2Inworld TTS1529
- 3MiniMax Speech 2.8 HD1529
- 4Deepdub eTTS 3.21533
- 5Lightning v3.1 Pro1525
TTS Arena · 2026-11-06
Transcription accuracy
Our picks- 1zoom/scribe_v2.1_pro4.87%
- 2OtterAI/speech-v3.15.25%
- 3p99lab/audrey-0.75.67%
- 4microsoft/azure-speech-07-20265.7%
- 5modulate/multilingual5.89%
Open ASR leaderboard, word error rate, lower is better · 2026-11-06
Building websites and apps
Our picks- 1Claude Opus 5.51810
- 2GPT-6 Astra1773
- 3Claude Sonnet 5.51770
- 4Gemini 4 Carbon1768
- 5GPT-6.1 Sol1751
LMArena WebDev · 2026-10-28
Search answers
Our picks- 1gpt-5.6-sol-xhigh1261
- 2qwen3.8-max1261
- 3claude-opus-4-6-search1251
- 4gpt-5.5-search1247
- 5claude-fable-5.1-high1246
LMArena search · 2026-10-13
Quick answers.
What is the smartest AI model right now?
Gemini 4 Carbon from Google leads the LMArena text leaderboard with a score of 1540, as of 2026-10-28. Scores close together are a statistical tie.
What is the best open-source AI model?
MiMo-V2.6-Pro is the highest-ranked open model (1491). You can download it, but the full version needs a server; smaller versions run on a home computer.
What is the LMArena score?
People ask two anonymous models the same question and vote for the better answer. Millions of votes become a score like a chess rating. Higher is better.
What does $ per 1M tokens mean?
What developers pay to use the model through its API. A million tokens is roughly 750,000 words. Chat apps charge a monthly plan instead.