Aitor Ormazabal (@aormazabalo) - Twitter Profili

Aitor Ormazabal retweetledi

Reka@RekaAILabs·12 Mar

98ms time to first token (faster than human visual reaction time), built for agentic workflows. 65% faster throughput compared to leading 8B models. Reka Edge is a 7B VLM built for latency-sensitive apps: real-time video analysis, agentic workflows, on-device deployment reka.ai/news/reka-edge…

English

2

8

43

2.7K

Aitor Ormazabal retweetledi

Reka@RekaAILabs·11 Mar

Meet Reka Edge – Our next-generation vision language model for physical AI. Uses 3x fewer input tokens and achieves 65% faster throughput compared to leading 8B models. Image understanding, video analysis, object detection, and tool use. Built for Action. Fast enough for production, deployable anywhere. Read more: reka.ai/news/reka-edge…

English

8

46

189

51.3K

Aitor Ormazabal@aormazabalo·25 Eyl

Check out parallel thinking on Reka Research for significant performance boosts at a negligible latency cost!

Reka@RekaAILabs

Introducing Parallel Thinking for Reka Research! Instead of one line of reasoning, we explore multiple paths in parallel, then resolve the best answer. Big accuracy gains on Research-Eval (+4.2) and SimpleQA (+3.5). Now live in the Reka Research API!

English

0

2

198

Aitor Ormazabal retweetledi

Reka@RekaAILabs·16 Eyl

Introducing Reka Speech: an efficient and accurate transcription & translation model. 🗣️ On modern GPUs (e.g. H100), it runs 8x–35x faster than existing solutions for batch processing.

English

10

28

141

25.2K

Aitor Ormazabal retweetledi

Reka@RekaAILabs·29 Ağu

🚨 New benchmark release 🚨 We're introducing Research-Eval: a diverse, high-quality benchmark for evaluating search-augmented LLMs 👉 Blogpost: reka.ai/news/introduci… 👉 Dataset + code: github.com/reka-ai/resear… 🧵 Here's why this matters:

English

7

24

91

26.4K

Aitor Ormazabal@aormazabalo·21 Tem

@YiTayML Super cool work Yi, congrats!

English

1

0

1

361

Yi Tay@YiTayML·21 Tem

First official Gold medal at IMO from DeepMind🥇 with Gemini Deep Think. A general purpose text-in text-out model achieving gold medal is something quite unthinkable just about one year ago and here we are! The frontier of AI is incredibly exciting! Happy to have co-led / captained model training! 😎 A fun fact was that I got roped into this effort thinking it was going to be a fun little side quest, little did I know that it turned out to be such a huge breakthrough. And yes, the results are certified by the IMO! 😉

Google DeepMind@GoogleDeepMind

An advanced version of Gemini with Deep Think has officially achieved gold medal-level performance at the International Mathematical Olympiad. 🥇 It solved 5️⃣ out of 6️⃣ exceptionally difficult problems, involving algebra, combinatorics, geometry and number theory. Here’s how 🧵

English

29

30

524

90.3K

Aitor Ormazabal retweetledi

Sharath Raparthy@sharathraparthy·18 Tem

“Apple looses key AI leaders to Meta” I discovered this while doing the live demo of Reka Research 😂 Go watch the video and play with our agent

Reka@RekaAILabs

Reka Research is our AI agent that scours the web to answer your toughest questions. Ready to unlock its full potential? Learn directly from the team who built it!

English

0

3

27

1.8K

Aitor Ormazabal retweetledi

Sharath Raparthy@sharathraparthy·16 Tem

looking for an agent which excels at questions that require dozens of sources and delivers accurate responses with reasoning traces in a few minutes? Reka Research is here for you. start building today: docs.reka.ai/quick-start

English

0

5

18

1.4K

Aitor Ormazabal@aormazabalo·14 Tem

@eagirre @RekaAILabs @Hitz_zentroa Milesker eneko!

Euskara

0

1

47

Eneko Agirre @eagirre.bsky.social@eagirre·14 Tem

Among the impressive developments at Reka, I'm particularly thankful to @RekaAILabs for open sourcing the much needed quantization library. Congrats @aormazabalo for the great work! We are very lucky for having you co-supervising PhDs at @Hitz_zentroa

Aitor Ormazabal@aormazabalo

For our third day of releases we are open sourcing some of our building blocks! I'm particularly happy to be open-sourcing RekaQuant 🗜️, part of our internal quantization stack that I led last year. Short thread on our approach to quantization 🧵1/n

English

1

13

634

Aitor Ormazabal@aormazabalo·11 Tem

In our experience, this approach significantly outperforms QAT or LDLQ alone for low bitwidths. Check out RekaQuant: github.com/reka-ai/rekaqu… Read the blog post: reka.ai/news/reka-quan… Check out Flash 3.1: huggingface.co/RekaAI/reka-fl… And its 3.5 bit quant: huggingface.co/RekaAI/reka-fl…

English

0

3

209

Aitor Ormazabal@aormazabalo·11 Tem

Instead, RekaQuant quantizes tensors one by one progressively through training, and does direct self-distillation for the unquantized tensors in between. This allows us to use any expensive quantization approach (LDLQ!), and is highly effective.

English

2

0

4

248

Aitor Ormazabal@aormazabalo·11 Tem

For our third day of releases we are open sourcing some of our building blocks! I'm particularly happy to be open-sourcing RekaQuant 🗜️, part of our internal quantization stack that I led last year. Short thread on our approach to quantization 🧵1/n

Reka@RekaAILabs

📢 We are open sourcing ⚡Reka Flash 3.1⚡ and 🗜️Reka Quant🗜️. Reka Flash 3.1 is a much improved version of Reka Flash 3 that stands out on coding due to significant advances in our RL stack. 👩‍💻👨‍💻 Reka Quant is our state-of-the-art quantization technology. It achieves near-lossless compression of Reka Flash 3.1 to 3.5 bits. 💻

English

1

10

35

6.6K

Aitor Ormazabal retweetledi

Mikel Artetxe@artetxem·10 Tem

We are open sourcing Reka Flash 3.1, our new 21B reasoning model! Big gains over the previous version, go build with it 🔥🔥🔥 And... we are also open sourcing Reka Quant, our new quantization method achieving SOTA compression to 3.5 bits!

Reka@RekaAILabs

📢 We are open sourcing ⚡Reka Flash 3.1⚡ and 🗜️Reka Quant🗜️. Reka Flash 3.1 is a much improved version of Reka Flash 3 that stands out on coding due to significant advances in our RL stack. 👩‍💻👨‍💻 Reka Quant is our state-of-the-art quantization technology. It achieves near-lossless compression of Reka Flash 3.1 to 3.5 bits. 💻

English

1

7

49

2.5K

Aitor Ormazabal retweetledi

Sharath Raparthy@sharathraparthy·10 Tem

🎉🚀 Release Day 3: It’s #OpenSource time! We’re pleased to open-source Reka Flash 3.1, our coding model achieving state-of-the-art performance on LiveCodebench and other benchmarks. 📈🔍 🛠️✨ We’re also opensourcing Reka Quant, our cutting-edge quantization tech that compresses Reka Flash 3.1 to just 3.5 bits with near-lossless fidelity🪄 Congrats to the team (@aormazabalo , Xiaonan and co) who worked on this.

Reka@RekaAILabs

📢 We are open sourcing ⚡Reka Flash 3.1⚡ and 🗜️Reka Quant🗜️. Reka Flash 3.1 is a much improved version of Reka Flash 3 that stands out on coding due to significant advances in our RL stack. 👩‍💻👨‍💻 Reka Quant is our state-of-the-art quantization technology. It achieves near-lossless compression of Reka Flash 3.1 to 3.5 bits. 💻

English

0

2

7

499

Aitor Ormazabal retweetledi

Reka@RekaAILabs·10 Tem

📢 We are open sourcing ⚡Reka Flash 3.1⚡ and 🗜️Reka Quant🗜️. Reka Flash 3.1 is a much improved version of Reka Flash 3 that stands out on coding due to significant advances in our RL stack. 👩‍💻👨‍💻 Reka Quant is our state-of-the-art quantization technology. It achieves near-lossless compression of Reka Flash 3.1 to 3.5 bits. 💻

English

12

50

206

735.5K

Aitor Ormazabal

Keşfet