Categories
AI News iRaluca

JetBrains Retrained Mellum, and Its Agent Coding Score Went From 2 to 47

JetBrains’ open Mellum2.1 uses 2.5B active parameters. New training alone lifted its SWE-bench Verified score from 2% to 47%, self-reported.

Categories
AI News iRaluca

Claude Haiku 5.5 Costs 90% Less, Until Your Prompt Passes 100k Tokens

Anthropic’s Haiku 5.5 adds a 1M context and effort controls. It’s 90% cheaper than Haiku 4.5 below 100k tokens, 50% cheaper above.

Categories
AI News iRaluca

Reflection Named Its 501B Open Model Beam. The Weights Aren’t Out Yet.

Reflection AI announced Beam, a 501B open-weight model with 23B active. Apache 2.0 is promised; the weights are due later this month.

Categories
AI News

Aleph Alpha’s Kolibri Runs 78B Parameters on a 3.46B Budget

Aleph Alpha released Kolibri, an Apache-licensed German-English model: 78B parameters, 3.46B active per token, trained to hold back when unsure.

Categories
AI News

Decision Models Go Open: AI That Answers by Choosing, Not Writing

Cloudflare open-sourced Clef and llama.cpp is adding support: models that pick from your options instead of generating text.

Categories
AI News iRaluca

A Student Taught Qwen3.8 to Mean “Medium” When It Says Medium

A solo developer fine-tuned Qwen3.8-27B so low, medium and xhigh effort settings behave in order. Self-reported gains, one seed, Apache 2.0.

Categories
AI News iRaluca

Naive AI Didn’t Pretrain Its Model. It Adopted Xiaomi’s.

A seven-month-old Beijing startup released a 309B open-weight model — built on Xiaomi’s MiMo-V2.5, with no full-attention layers anywhere.

Categories
AI News iRaluca

The Model Built Purely to Guess What Another Model Will Say

Liquid AI’s new 280M draft model speeds up its vision model up to 3.13x — but only decoding, and only free under a $10M revenue cap.

Categories
AI News iRaluca

Transformers Can Now Load and Run llama.cpp’s GGUF Files

Hugging Face’s Transformers library can now load and run llama.cpp’s GGUF quantized models directly, matching its speed on Apple Silicon.

Categories
AI News iRaluca

A Slowdown Essay, Then Two Frontier Models Ten Days Later

Anthropic’s CEO called for slowing AI capability gains on 12 September. Ten days later, both labs shipped faster, cheaper frontier models.