Tag
一位评测者在 7900 XTX 上花 16 小时对比了名字夸张的 Qwen3.8-27B-Coder390-EfficientThink 变体与官方原版等模型，发现准确率无明显差异，主要差别在响应速度，最终推荐了更快的 ThinkingCap 版本。
OpenCode usage data shows a mystery AI model called fledge-alpha ranked #36 with 5.6B tokens, prompting speculation about which lab is behind it since its origin remains unknown.
Anthropic released a research report claiming that 智谱AI's open-source model GLM-5.3 can independently write usable cyberattack programs, with capabilities approaching Claude Mythos Preview, and that its security defenses are easily bypassed.
A tweet exposing past fraud where Reflection 70B was falsely presented as an open-source model but served Claude API, leading to bans in the AI community.
PixelJev is introduced as a native-image decision interface using small open multimodal models to map images, instructions, and candidate sets to structured choices. The study demonstrates adaptation improves accuracy on benchmarks like Pets and highlights challenges in calibration and generalization.
This post discusses the recent release of the K2-Horizon-MoVA-36B-A4B MoE model and asks about its current usage and potential use cases, providing a link to model comparisons.
GLM-5.3 achieves 310 tokens per second on Databricks inference, leading in both speed and latency, and is the strongest open-source model for coding, competitive with Fable 5 and Opus 4.8.
The author benchmarked Ox Alpha on SWE-bench Verified Mini, achieving a 96% resolution rate, but raises concerns about the score's validity due to data contamination, small sample size, and non-comparable baselines.
This paper releases an open eleven-structure cardiac CT statistical shape model and compares shape completion methods, demonstrating that a closed-form conditional-Gaussian estimator outperforms deep learning approaches like variational autoencoders.