AI models league table
Which AI model is best right now, in plain English.
Top three models
What do you want help with?
The overall ranking across all tasks
New and not ranked yet: Gemini 4 Hydrogen, Claude Fable 5.2, GPT-6.1 Terra, Kimi K3.1 Pro. They'll join the table once there's enough independent data. See Not fully scored
Showing 6 of 79 models
Click any model for full details and visitor ratings.
| Rank | Lab | Model | Released | Score | Cost to use | Visitor rating | Details |
|---|---|---|---|---|---|---|---|
| 20 | Z.ai | Good for: health questions, science and research, coding | 7.5/10 | Not confirmed | – | ||
| 32 | Z.ai | Good for: coding, maths, health questions | 7.1/10 | Free plan | – | ||
| 38 | Z.ai | Good for: health questions | 6.8/10 | Not confirmed | – | ||
| 45 | Z.ai | Good for: health questions, coding | 6.3/10 | Not confirmed | – | ||
| 55 | Z.ai | Good for: health questions, coding | 5.8/10 | Not confirmed | – | ||
| 67 | Z.ai | Good for: health questions | 5.4/10 | Not confirmed | – |
- 20
Z.ai ·
Not confirmedGood for: health questions, science and research, coding
Visitor rating–7.5/10 - 32
Z.ai ·
Free planGood for: coding, maths, health questions
Visitor rating–7.1/10 - 38
Z.ai ·
Not confirmedGood for: health questions
Visitor rating–6.8/10 - 45
Z.ai ·
Not confirmedGood for: health questions, coding
Visitor rating–6.3/10 - 55
Z.ai ·
Not confirmedGood for: health questions, coding
Visitor rating–5.8/10 - 67
Z.ai ·
Not confirmedGood for: health questions
Visitor rating–5.4/10
How to read this
Each model gets a score out of 10. 10 is the best model today; 9 is very close behind and anything under 7 is well behind. We also show whether you can use it free, and what it is good at. Some models are too new to have enough Arena votes; they move into the main table once Arena lists them. How we score
Updated 16 Nov 2026 · 79 models
Safety and ease-of-use checks are not part of the score yet.
Not fully scored
These models are public but we don't have a standard test result for them yet, so we can't give them a fair score. They are not ranked.
Gemini 4 HydrogenNo benchmark score yetArena score 1490 from 4,741 votesGoogleFree planKimi K3.1 ProNo benchmark score yetArena score 1485 from 3,995 votesMoonshotFree plan
MiMo V2.6 ProNo benchmark score yetArena score 1480 from 4,064 votesXiaomiFree plan
ERNIE 5.1No benchmark score yetArena score 1467 from 39,018 votesBaiduNot confirmed
Dola Seed 2.0 ProNo benchmark score yetArena score 1456 from 78,156 votesByteDanceNot confirmed
- Hy3No benchmark score yetArena score 1454 from 15,178 votesTencentNot confirmed
- Step 5 PreviewNo benchmark score yetArena score 1449 from 12,345 votesStepFunDevelopers only
MiMo V2.6 FlashNo benchmark score yetArena score 1448 from 23,252 votesXiaomiFree plan
Claude Fable 5.2No Arena votes yetBenchmark score 173.4 (Epoch Capabilities Index)AnthropicPaid plan
GPT-6.1 TerraNo Arena votes yetBenchmark score 167.7 (Epoch Capabilities Index)OpenAIPaid plan
Muse Spark OpenNo Arena votes yetBenchmark score 155.4 (Epoch Capabilities Index)MetaNot confirmed
Arena leaderboard data, CC BY 4.0, adapted by Haystak. Benchmark data: Epoch AI Capabilities Index, CC BY 4.0. Prices: Artificial Analysis and provider pages.