Evren Ozkip

GPT-6.1 Sol arrived just one week after GPT-6 Sol. Its Artificial Analysis Intelligence Index score increased by 4 points. At max reasoning, its average cost per benchmark task is less than a quarter of Astra’s: $0.72 vs. $3.26.

20 minutes ago | [YT] | 0

Evren Ozkip

Claude Sonnet 5.5 is here! 🚀

30% faster and up to 30% cheaper than Sonnet 5.

Same token pricing. Fewer tokens needed.

It also scores close to Opus 5.5 on the Artificial Analysis Intelligence Index: 56 vs. 58.

1 day ago | [YT] | 3

Evren Ozkip

AI can evaluate another AI’s answer. 🤖

For example: “Does this answer match the provided document?”

JEV evaluates it first. If its confidence is high enough, the decision is accepted. If not, a stronger model takes over.

The goal: lower evaluation costs by using powerful models only when needed.

1 day ago | [YT] | 3

Evren Ozkip

Claude Opus 5.5 is here! 🚀

Anthropic’s new Opus model looks especially strong for AI agents, coding, research, and complex tasks.

Here are the highlights:
⚡ 30%+ faster than Opus 5
💰 Around 40% lower cost to run
📥 $4 per 1M input tokens
📤 $20 per 1M output tokens
🧠 Cache reads are 60% cheaper

Some early results are impressive:

🔹 Airbnb completed a 680,000-line code migration in less than a day.
🔹 In Deloitte’s test, Opus 5.5 found 72% of known bugs, compared with 56% for Opus 5.
🔹 On BigFinance, Opus 5.5 at low effort beat Opus 5 at high effort while using about 60% fewer output tokens.

The AI model race is not only about which model is smarter anymore.


For AI agents, speed, cost, token efficiency, and reliability matter too.


What do you think about Opus 5.5? 👇

1 week ago | [YT] | 3

Evren Ozkip

Meet Qwen3.8-Flash ⚡

Qwen just introduced a new multimodal MoE model, and it also gives us an early look at the architecture being explored for Qwen4.

Highlights:

• 125B parameters
• Only 6B active per token
• 1/9 the training cost of Qwen3.7-Plus
• 262K native context
• Up to 1M with YaRN
• Open-weight

I created a short carousel to summarize the key points.

Skip the Carousel

1 2 3 4 5 6 7

1 month ago | [YT] | 5

Evren Ozkip

I’m happy to share that I participated in the manuscript review of Engineering Agent Systems for Manning Publications. 📘

I reviewed the book and shared my technical feedback.

It was a great opportunity to learn more about AI agent systems and contribute to a new book in this field.

Thank you to Manning Publications and Angelina Lazukić for the opportunity. 🙏

1 month ago | [YT] | 0

Evren Ozkip

New video is live!

Explore the Hugging Face ecosystem and learn how to build end-to-end AI apps with LLMs.

https://youtu.be/r477gqo5Oy0

9 months ago | [YT] | 7

Evren Ozkip

I’m excited that OpenAI has finally released two open-source models under the Apache 2.0 license.

- The large model runs on a single H100 GPU
- The small model fits within just 16GB of memory

Both are accessible through open-source frameworks like Transformers and Ollama.

I can't wait to build AI apps with these models.

1 year ago | [YT] | 3

Evren Ozkip

AI agents are programs that can autonomously implement a task on behalf of a user. Which AI tool frameworks do you use the most?

1 year ago | [YT] | 2

Evren Ozkip

Meta's new large language model Llama 3 is out. The good news is that you can use this open-source model for free in your own projects.

Llama 3 8B is better than Gemma 7B and Mistral 7B. Llama 3 70B is better than Gemini Pro and Claude 3 Sonnet.

For more information: llama.meta.com/docs/get-started

Skip the Carousel

1 2 3

2 years ago | [YT] | 2