Where AI meets the real world. Formerly LMArena. We measure and advance the frontier of AI through community-driven evaluation. We’re hiring → arena.ai/jobs
US
- Tweets 3,420
- Following 216
- Followers 173,895
- Likes 2,510
Introducing Agent Mode: Agentic AI is now measured in the Arena.
Agent Mode can do deep research, create reports, generate images, build websites, debug code, and more.
It completes more complex tasks by using tools like web search, bash in a sandbox environment, image generation, file writing, and asking follow-up questions.
Frontier models are waiting for you in Agent Mode to take on real-world tasks. GPT-5.5, Claude Opus 4.7, Gemini 3.1 Pro, and top open models. Test them yourself.
54
56
538
291,328
Arena.ai retweeted
I spent a LOT of time through the hardest 3D prompts at Fable, it is a 45 min video, but I have 60+ very cool demos for you. Also prompts in the next post.
piped.video/watch?v=rTc2_-1K…
45
112
1,316
334,725
Arena.ai retweeted
Fable - same prompt, different reasoning efforts - 12 mins vs nearly 2 hours between 'low' and 'max'
49
19
958
235,499
Claude Fable 5 is back - it might be the most impressive model we've ever seen.
@petergostev ran it through 60+ of the hardest 3D generations: modern cities, art, world wonders, and more.
The results are incredible. Fable is clearly in a league of its own. Watch youtube here:
piped.video/watch?v=rTc2_-1K…
28
40
643
189,327
Now that Fable 5 is back in the Arena, check out this montage of a few of the prompts we put it through.
Find it in Battle Mode and Agent Mode, and start testing it for yourself. Official leaderboard scores coming soon.
piped.video/watch?v=euiewern…
1
5
37
11,515
Fable 5 is back in the Arena!
When it first debuted, Fable 5 ranked #1 in Agent Arena: our benchmark for real-world, long-horizon agentic performance. Agent Arena evaluates models on millions of real tasks submitted by a global community of users, with access to web search, filesystem, and terminal tools to complete complex workflows. Rather than raw win rates, the leaderboard uses causal tracing to measure how much each model's outcomes outperform the average model.
@AnthropicAI's latest model is also live in Text, Vision, Document, and Code Arena: Frontend.
Come give it a try in Battle Mode and Agent Mode and contribute to the leaderboards.
23
19
351
165,588
Now that Fable 5 is back in the Arena, check out this montage of a few of the prompts we put it through.
Find it in Battle Mode and Agent Mode, and start testing it for yourself. Official leaderboard scores coming soon.
piped.video/watch?v=euiewern…
1
5
37
11,515
Now that Fable 5 is back on Arena, watch @petergostev put the re-deployed model by @anthropicAI through 60+ of the most complex 3D generations, mini-games, and world-building tests.
Watch on YouTube: piped.video/watch?v=rTc2_-1K…
1
12
9,605
The community has been asking how Claude Fable 5 compares before vs. after its latest re-deployment.
We collected thousands of votes on the new endpoint across Arenas - Text, Vision, Document, Code, and Agent - and here’s an early score preview.
So far, scores look mostly consistent before and after re-deployment. Fable 5 remains at the frontier across Text, Document, Vision, and Code Arena: Frontend. The ~20-point drop in Frontend is still within the confidence interval as scores continue to stabilize.
We’ll share more insights as more data comes in across all arenas - stay tuned!
Fable 5 is back in the Arena!
When it first debuted, Fable 5 ranked #1 in Agent Arena: our benchmark for real-world, long-horizon agentic performance. Agent Arena evaluates models on millions of real tasks submitted by a global community of users, with access to web search, filesystem, and terminal tools to complete complex workflows. Rather than raw win rates, the leaderboard uses causal tracing to measure how much each model's outcomes outperform the average model.
@AnthropicAI's latest model is also live in Text, Vision, Document, and Code Arena: Frontend.
Come give it a try in Battle Mode and Agent Mode and contribute to the leaderboards.
50
100
1,160
181,656
Test out Claude Fable 5 in Battle Mode and Agent Mode across modalities and contribute your votes. Final scores for the leaderboard coming soon: arena.ai
3
1
29
7,971
See how Claude Fable 5 originally stacked up on the leaderboards at: arena.ai/leaderboard
1
25
6,758
Arena.ai retweeted
So excited to share that Fullstack Code Arena is officially out. It was so fun getting to build this with the team :)
As AI systems continue to push the frontier, our evaluations need to advance alongside them, measuring beyond frontend prototypes. Fullstack brings a new way to holistically measure AI Agents in building the kind of software people use daily.
Can’t wait for you to try it out and hear what you think!
Code Arena just leveled up with fullstack capabilities 🚀
Introducing the new Fullstack Code Arena. We’re moving beyond frontend prototypes to fullstack development complete with databases, API keys, and fast deployments. Build, iterate, and ship real-world software — all in one place.
Models now act as agents in the Code Arena, using structured tool calls to plan, execute, and refine in real time with real world tasks.
Read more about it in the thread 🧵
2
1
26
7,379
With Fullstack, we now support building more complex apps that require database or backend functionality. Code Arena now powers building a whole new suite of apps you couldn’t before, including:
- Apps that require a sign up or login flow, such as admin portals or dashboards
- Apps that use an API key, like an AI chatbot
- Apps that store a user’s progress for them to come back to, such as an online education or teaching app
This changes Code Arena from more of an experimental prototyping environment to a daily-use tool where real work gets done. The votes will power a future Fullstack leaderboard.
Click through to our YouTube to watch the full walkthrough of Fullstack Code Arena:
piped.video/Eu-gcfuxGn8
2
2
21
7,021
Top 20 Startups by Web Traffic founded since 2020
1. DeepSeek
2. Perplexity
3. Suno
4. Polymarket
5. Gamma
6. ElevenLabs
7. Lovable
8. Arena
9. xAI
10. Supabase
11. Manus
12. Higgsfield
13. Cursor
14. Fanvue
15. OpenRouter
16. GPTZero
17. Genspark
18. ShopMy
19. Venice
20. Whop
Some interesting observations:
— Only 25% were not AI: Polymarket, Supabase, Fanvue, ShopMy, Whop
— 20% were acquired
— Startups that didn't surprisingly didn't make the cut: Kalshi (founded 2018), Mistral (10M), OpenEvidence (11.4M), Cognition
— All but 2 are unicorns (GPTZero, Fanvue), 7 decacorns, but there's no clear correlation between traffic and valuation
69
87
651
177,150
Code Arena just leveled up with fullstack capabilities 🚀
Introducing the new Fullstack Code Arena. We’re moving beyond frontend prototypes to fullstack development complete with databases, API keys, and fast deployments. Build, iterate, and ship real-world software — all in one place.
Models now act as agents in the Code Arena, using structured tool calls to plan, execute, and refine in real time with real world tasks.
Read more about it in the thread 🧵
11
17
159
28,914
Code Arena now goes beyond frontend prototypes. Read more about all the new fullstack features in our blog:
arena.ai/blog/fullstack-code…
1
1
10
3,907
Get testing and building with the new Fullstack Code Arena at: arena.ai/code
1
5
3,244
Claude Sonnet 5 (Thinking) has landed #6 for Code Arena: Frontend.
The latest Sonnet scores +29 points higher than its predecessor, Sonnet 4.6, on agentic frontend webdev tasks, and outscores Opus 4.6 (Thinking) by +9 points. It also ranks widely across other modalities:
- #11 in Document Arena
- #17 in Search Arena
- #21 in Vision Arena
- #32 in Text Arena
More details on how Sonnet 5 differs over 4.6 in thread.
Congrats to @AnthropicAI on this release.
Introducing Claude Sonnet 5, our most agentic Sonnet yet.
It makes plans, uses tools like browsers and terminals, and runs autonomously at a level that just a few months ago required larger and more expensive models.
44
70
792
93,442
Claude Sonnet 5 (Thinking) landed #32 overall for Text. Let's dig into why.
In the Text Arena, you can filter by category and occupational domain. Notably, Sonnet 5 outranks 4.6 on Expert-level prompts, and holds steady in categories like Math, Writing, Literature & Language, and Life, Physical & Social Science - but drops in rank across most other categories.
6
4
37
5,543
Learn more about Agent Arena’s causal tracing methodology with Evan:
piped.video/watch?v=0-qw5Emg…
1
6
7,198
Fable 5 is now in Agent Mode - we’ll see if it holds the lead as new traces come in. arena.ai/agent
1
11
4,665