DeepSeek V4 Flash vs V3.2: Users Report Regression
Early users of DeepSeek V4 Flash report it underperforms its predecessor V3.2, sparking debate about speed-quality trade…
📂 LLM News
Latest articles in LLM News
LG AI Research Launches EXAONE 4.0 Multimodal Model
LG AI Research unveils EXAONE 4.0, a multimodal foundation model bringing vision-language capabilities to its enterprise…
Kakao Brain Launches Open-Source Korean Vision-Language AI
Kakao Brain releases an open-source vision-language model optimized for Korean, expanding multilingual AI capabilities b…
NTT Builds Lightweight Japanese LLM Matching GPT-4
Japan's NTT develops a compact Japanese language model that achieves GPT-4-class performance with significantly fewer pa…
Preferred Networks Launches PLaMo-2 for Japan
Japanese AI firm Preferred Networks unveils PLaMo-2, a large language model built for enterprise use across Japanese ind…
AI21 Labs Launches Jamba 2 Hybrid SSM-Transformer
AI21 Labs unveils Jamba 2, a next-gen model blending State Space Models with Transformer attention for faster, more effi…
Stability AI Launches Creative Writing LLM
Stability AI releases a new open-weight language model specifically tuned for fiction, poetry, and long-form creative wr…
Claude 4 Opus Smashes Graduate-Level Benchmark Records
Anthropic's Claude 4 Opus sets new state-of-the-art scores on GPQA and other graduate-level reasoning benchmarks, outpac…
Meta Launches Llama 4 Maverick: 400B Open Model
Meta releases Llama 4 Maverick, a 400B-parameter mixture-of-experts model under an open license, challenging GPT-4o and …
Mistral AI Launches Codestral 2.0 for Enterprise
Mistral AI unveils Codestral 2.0, a specialized code generation model targeting enterprise developers with improved accu…
Cohere Command R+ Boosts RAG for Enterprise AI
Cohere launches Command R+ with a significantly enhanced retrieval-augmented generation pipeline, targeting enterprise c…
Google Gemini 2.5 Ultra Hits Enterprise Preview
Google rolls out Gemini 2.5 Ultra in limited enterprise preview, targeting large-scale business AI deployments with enha…