93-Model Benchmark Over 1.5 Years: Distillation + Quantization + RL Hurts General AI
Afinetheorem · x · 2026-10-19
Researcher Afinetheorem ran a personal image + logic benchmark (HardGeoBench, an image-based geolocation eval) across 93 models over 1.5 years and argues that distillation plus quantization plus task-specific RL causes big harms to general AI capability.
- Worst model ever: DeepSeek v4.1 Flash 1017
- The four worst also include GPT 6 Luna, Mimi 2.6 Pro, and Qwen 3.8 27b
- The leaderboard shows DeepSeek v4.1-flash-1017 at just 3.8% accuracy within 10 miles and 3201.4 miles average distance error, versus 10–17% for Gemini, Claude, and Grok models
More from Models
- Hazy Research's open-source NCM generates text 31x faster via coarse-to-fine concept modeling — simran_s_arora · 2026-10-20
- Blogger claims DeepSeek's unreleased model nearly matches Opus at design, shares one-prompt demo — sven_ai · 2026-10-20
- Fingerprint test flags DeepSeek ds-4.1-flash as likely GPT-6-astra — sven_ai · 2026-10-20
- Intelligence Saturation: 10-Point Model Edge Now Buys Just 20% More Developer Traffic — khademinori · 2026-10-20
- MiniMax M3.1 Flash Hits 73.3 on BuildingBench, Reshaping the Cost-Quality Frontier — Lianhuiq · 2026-10-20
- User: Codex burns half of the day's usage limits within an hour after banked reset — ssh4net · 2026-10-20