EN · International edition · 28/10/2026
Inferama daily · 28/10/2026
The most important AI developments since the previous newsletter, selected for editorial relevance.
News coverage period: 27/10/2026 08:30 → 28/10/2026 08:30 (Europe/Madrid).
Recursive Flint: The 40B Local Coding Model That Runs on a Laptop
Recursive Superintelligence has released Flint, an open-weight model with 39.8 billion parameters designed for local code inference. With only 2.5 billion active parameters per token, the company claims speeds exceeding 125 tokens per second on Apple Silicon laptops.
Read article →
OpenAI presents three scientific results for academic evaluation
OpenAI has announced results related to the Kakeya conjecture, the Exponential Time Hypothesis and quantum chromodynamics. The company says academic researchers will examine and build on them; the available evidence does not establish that they have been independently validated.
Read article →
Genie 3 generates interactive worlds in real time, but sessions remain brief
Google DeepMind presented Genie 3 as a world model that turns text prompts into navigable, interactive environments. The research preview runs at 24 frames per second and 720p, maintains consistency for a few minutes, and still has limits on actions, duration, and geographic accuracy.
Read article →
InfraDev evaluates coding agents on 36 AI infrastructure changes
The benchmark draws on tasks based on real changes to Megatron-LM, vLLM, and verl. In the preprint, the best result among six frontier models was solving 38.9% of the tasks.
Read article →
Liquid AI introduces LFM3, a family of multimodal models for specialization
LFM3 brings together text, vision, audio, and decision models built from a shared foundation. Liquid AI says the model weights are available and each model can be fine-tuned on a single GPU; the published performance figures are the company’s own evaluations, not independent verification.
Read article →
Maintaining Code Repositories with AI: The Verification Protocol Before Accepting Changes
A coding agent’s autonomy depends on authorised scope and available verification as well as model quality. This guide explains how to combine tests, review and permissions to reduce regressions and data exposure without presenting those controls as an absolute guarantee.
Read article →
Operational Viability Matrix: Deciding Between Automating, Assisting, or Staying Manual
A decision guide focused on real execution conditions: latency, error rates, review costs, and data availability. Learn to calculate the break-even point where AI automation becomes less cost-effective than human assistance or manual processes, avoiding complex adoptions when deterministic simplicity is superior.
Read article →