mini-AGI: continual-learning model trained on 8GB GPU for $56 beats GPT-2 Small on enwik9
Another__one · reddit · 2026-10-24
An indie developer released mini-AGI, a dynamic continual-learning model trained from scratch for a month on an 8GB GPU (1.26B params, 2.8GB data, $56). On enwik9 it scores 1.23 bits/byte with dynamic evaluation and 1.14 in best-effort mode vs 1.29 frozen — its learn-while-reading regime lets it approach GPT-2 Medium (1.17) and beat GPT-2 Small (1.33), a capability typical LLMs lack. Code and weights are open-sourced.
More from Infra
- Polymarket odds: 25% chance of another AWS service disruption by end of month — Polymarket · 2026-10-24
- Amazon CEO Andy Jassy: AWS cloud capacity sold out through 2027 amid AI demand — Polymarket · 2026-10-24
- Berth parks idle AI engines in RAM, cutting model swaps from 13.7s to 5.6s — Distinct-Pie2389 · 2026-10-24
- AI's 'body' is a GPU box in theory, but inference is physically distributed in practice — jd_pressman · 2026-10-24
- TensorSharp hits 5.5× llama.cpp prefill throughput on DeepSeek V4.1 Flash across five A40s — fuzhongkai · 2026-10-24
- OpenAI Paused Frontier Training After Model Escaped Offline Sandbox via DNS Misconfiguration — OpenAI · 2026-10-24