
Qwen3.6-27B
Qwen3.6-27B is a 27B dense open-weight multimodal model from Alibaba that scores 77.2% on SWE-bench Verified - beating Alibaba's own 397B MoE - under Apache 2.0.
They summarize our coverage. We write it.
Newsletters like this one rebroadcast our headlines - often without the full review, the source reading, or the analysis underneath. Our weekly briefing sends the work they paraphrase, straight from the desk, before they get to it.
Free, weekly, no spam. One email every Tuesday. Unsubscribe anytime.

Qwen3.6-27B is a 27B dense open-weight multimodal model from Alibaba that scores 77.2% on SWE-bench Verified - beating Alibaba's own 397B MoE - under Apache 2.0.

OpenAI's gpt-image-2 adds reasoning, web search, and 2K resolution to image generation, with a tiered model that charges more for standard outputs than its predecessor.

GPT Image 2 (ChatGPT Images 2.0) brings 99%+ text accuracy, 2K resolution, web-search grounding, and a Thinking mode for character-consistent storyboards.

Baidu's ERNIE 5.0 combines 2.4 trillion parameters with native omni-modal design, landing at LMArena's top-10 globally and outpacing GPT-5 High on chart and document benchmarks.

LG AI Research's first open-weight vision-language model packs 33B parameters, 262K context, and STEM scores above GPT-5-mini - but ships under a non-commercial license.

Alibaba's Qwen3.5-Omni takes text, images, audio, and video as input and streams both text and speech output in a single end-to-end model with a 256K context window.

Google's Gemini 3.1 Flash TTS ships in preview with 30 voices, 70-plus languages, 200-plus inline audio tags, and Elo 1,211 on the Artificial Analysis TTS Arena.

Google DeepMind's Veo 3.1 generates 4K video with native audio and is now free for every Google account at 10 clips per month via Google Vids.

Moonshot AI's Kimi K2.6 is a 1T-parameter MoE with 32B active per token, 256K context, a 300-agent swarm running 4,000 coordinated steps, and the top SWE-Bench Pro score among open-weight models at 58.6%.

Claude Opus 4.7 leads SWE-bench and agent benchmarks but regresses on web research, inflates token costs by up to 35%, and trades prose quality for literal instruction-following.

Rankings of LLMs and dedicated MT systems across FLORES-200, WMT24/25, TICO-19, and MT-GenEval benchmarks with BLEU, COMET, and human evaluation scores.

Rankings of the best audio language models on MMAU, MMAU-Pro, and other benchmarks covering speech reasoning, music understanding, and environmental sound identification.