Kimi K3.1 is out; it is the #1 OSS model. It places similarly to Opus 5, which was released 3 months ago. Thus, China is lagging the fronti...

Pretty interesting test-time scaling plot. Basically, you should use Astra Max for your hardest tasks. Then, for your next-hardest tasks,...

Muse Wonder Full is the #6 model on the Code Arena, close to Fable 5.1 in performance. Huge congrats to the Meta team! Truly a frontier mo...

Congratulations @carrieeeeet and team on the launch of Readable Intelligence! https://twitter.com/carrieeeeet/status/2112228826768142652

Reflection's Beam model is the #2 most aligned OSS model, and the #1 most aligned American OSS model. It is substantially ahead of Thinking...

Looks like Muse Spark Open is the #1 US OSS model. Meta is truly back in the saddle. Incredible comeback. Congrats @alexandr_wang @finkd ht...

This is great news. Max is better than any individual model, and 37% cheaper than Fable 5.1. You can try it today on @arena! https://twitter...

Check out the Arena Alignment Index: https://arena.ai/leaderboard/agent/alignment Thanks for having me @johncoogan @jordihays https://twitt...

Arena has raised a $200M Series B at a $3.1B valuation. We are using this momentum to evaluate agentic utility in real workflows from every...

Step 1: Make a router Step 2: Make a ton of noise on social media Step 3: Profit Step 4: Test the router Step 5: It is not on the pareto...

Stop tipping Claude $15 by saying "Thank you" at the end of a long conversation with an invalidated cache https://twitter.com/RashiShrivast1...

The chakras have realigned Google hits #1 on Text Arena Top-3 provider in Code Arena https://x.com/ml_angelopoulos/status/2105397462177644...

Yep. This is a toxic (and crazy) outlook. There are plenty of important problems that basic research will solve outside of labs. Much of th...

Was checking out Theranos' financings: Seed: DFJ A: Angels (Murdoch and others) B/C: ATA Ventures (tier-3 firm, now shut down) From C onwa...

When we started @arena as a project at UC Berkeley, we were ranking eight models. Today we evaluate the performance of over 500 models, with...

Man! The pace at which labs are hill-climbing on coding is extraordinary. In the last week, there have been 4 models in the #1 spot: - GPT-...

There's some leapfrogging happening between Anthropic and the most recent Chinese model release. Basically they're in a race for the #1 mo...

Gemini 3.8 flash is a big improvement over 3.7, especially in terms of user satisfaction. Double-digit lift in the percentage of explicit ...

If you are an academic studying the foundations of AI Evaluation, we'd love to work with you. https://twitter.com/arena/status/2094871997478...

This is the only primary-source interview on the METR report I have seen. This is one of the most important topics in AI this year, given th...