Alibaba's Qwen3.8-Max-0902 Tops Code Arena, Beats Claude Opus 5
Alibaba's refreshed Qwen3.8-Max-0902 model has taken the top spot on Arena.ai's Code Arena coding leaderboard, narrowly outscoring Anthropic's Claude Opus 5.
Read more →Large Language Modelsand News
Every AI News story tagged with both Large Language Models and News — the two topics side by side, updated as new articles publish.
14 articles
Anthropic Launches Claude Fable 5.1 and Mythos 5.1
Anthropic released Claude Fable 5.1 and Mythos 5.1 on September 1, cutting cache-read costs 75% and posting sharp benchmark gains over Fable 5.
Read more →Tencent Open-Sources Hy4 Preview, a 770B-Parameter Flagship Model
Tencent's Hunyuan team open-sourced Hy4 preview, a 770-billion-parameter model with a 1-million-token context window, joining China's fast-moving open-weight AI race.
Read more →Z.ai Unmasks Mystery 'Ox Alpha' Model as GLM-5.3-Flash
Chinese AI firm Z.ai confirmed its anonymous "Ox Alpha" model was GLM-5.3-Flash, releasing the open-weight, multimodal model under an MIT license.
Read more →Thomson Reuters Launches Thomson, Its Own Legal AI Model
Thomson Reuters launched Thomson, a proprietary AI model trained on its legal and news archives and built atop Alibaba's open-weight Qwen model, first powering CoCounsel Legal's document review.
Read more →AI Models Recover Only 3–15% of Research Ideas, Study Finds
A new benchmark testing seven frontier AI models found they could reconstruct a scientific paper's core idea from its pre-publication bibliography alone only 3-15% of the time.
Read more →Nvidia Releases Nemotron 3.5 Lightning to Cut AI Agent Costs
Nvidia has released Nemotron 3.5 Lightning, an open 30-billion-parameter AI agent model, plus NeMo Switchyard, a router that cuts task costs to about a third of using Opus 4.8 alone.
Read more →DeepSeek Raises API Prices Up to 12x With Peak-Hour Billing
DeepSeek will begin peak and off-peak API pricing on August 16, raising some DeepSeek-V4-Pro rates by up to 12 times as demand for its models surges.
Read more →DeepSeek Quietly Ships V4 Pro 0813 to Mixed Reviews
DeepSeek's flagship model exited preview on August 12 with sharp gains on coding and agent benchmarks, but only middling scores on general reasoning tests, and no formal announcement.
Read more →OpenAI's GPT-5.6-Cyber Reaches 'High' Cybersecurity Threshold
OpenAI released GPT-5.6-Cyber, a restricted-access model for vetted security researchers that it rates 'High' for cyber capability — one step below its 'Critical' threshold.
Read more →Thinking Machines Releases Inkling-Small, Outperforms Larger Model
Thinking Machines Lab's second open-weight model needs a quarter of its predecessor's active parameters yet edges past it on reasoning and coding benchmarks.
Read more →GPT-5.6 Sol Rewrote OpenAI's GPU Kernels, Cutting Serving Costs 20%
OpenAI says its GPT-5.6 Sol model autonomously rewrote the GPU kernels behind its inference stack, cutting serving costs 20% and helping fund an 80% price cut for the lightweight Luna tier.
Read more →Alibaba Unveils Qwen3.8-Max, a 2.4-Trillion-Parameter AI Model
Alibaba previewed Qwen3.8-Max, a 2.4-trillion-parameter multimodal model it calls second only to Anthropic's Claude, though no benchmarks have been published.
Read more →Fireworks AI Raises $1.5B Series D, Hits $17.5B Valuation
Fireworks AI closed a $1.5 billion Series D at a $17.5 billion valuation, saying its infrastructure now serves over 40 trillion tokens a day for specialized AI models.
Read more →