Qwen

Frontier intelligence now runs on a laptop. Qwen 3.8 27B tied GPT 5.6 Luna in the Artificial Analysis Intelligence Index rankings. Open source is not behind anymore. It is a year away from the top.

Qwen 3.8 27B ties GPT 5.6 Luna

My go-to AI stack for August 2026, everything here I’ve put 100+ hours on personally I use Fable 5 medium as the planning brain, it’s the sharpest model I’ve found for mapping out ideas ChatGPT 5.6 sol medium is the… 더 보기

Qwen 3.8 Max is out and it's already shaking things up. It scored 86.6 on TerminalBench, edging out Fable 5's 84.6, though GPT 5.6 still leads at 88.8. Pricing is wild at $2 per million input tokens and $6 per million… 더 보기

Qwen 3.8 Max TerminalBench score

Open-source models are quickly catching up to proprietary ones in agentic coding tasks. The release of the Ornith-1.0 family shows that a closed API isn't necessary for achieving state-of-the-art coding agent… 더 보기

Ornith-1.0 agentic coding LLM performance

GLM 5.2 has surged on the Artificial Analysis Coding Index, jumping from 43.4 to 50.7 , a gain of seven points in a single release. According to bridgemindai, these results align with what they're observing on… 더 보기

GLM 5.2 Artificial Analysis Coding Index

GLM 5.2 achieved the first perfect score on the BridgeBench BS benchmark, prompting plans for BridgeBench v3.

GLM 5.2 perfect score BridgeBench BS

Gemma 4 31B proves more intelligent than GPT 5 was. And I am running it locally on a MacBook. I just tested Gemma 4 31B and Qwen 3.6 35B on my $5,399 MacBook Pro M5 Max with 128GB unified memory. The speeds are insane.… 더 보기

Qwen 3.6 35B is achieving more than 100 tokens per second on my $5,399 MacBook Pro M5 Max. This counts as the best local AI model I have ever run. It uses 128GB of unified memory. There is no cloud involved. No API costs… 더 보기