104 Custom Tinker Models Analyzed: Spanning Reasoning, Safety, and Agents
davidthesong · reddit · 2026-07-30
A developer has cataloged 104 public AI projects built using Tinker, a post-training API developed by Thinking Machines.
These projects highlight Tinker's broad utility in enabling developers to customize AI for specific needs:
- Capability Enhancements: Improving reasoning, solving math problems, and advancing visual/creative abilities.
- Research & Automation: Studying AI safety and model behavior, training agents, and automating AI research workflows.
- Coding & Forecasting: Writing and optimizing code, forecasting financial markets, medicine, and global events.
Contributors include academic researchers, startups, large enterprises, and independent OSS developers, with fine-tuned model sizes ranging from 4B to 397B+ parameters.
Related event: 104 Projects Built on Tinker API Span Multiple AI Domains(2 posts)→
More from Models
- Testing Claude Opus 5: Minimal Prompts and Lightweight CLAUDE.MD Work Best — omarsar0 · 2026-07-30
- Claude Opus 5 Tops Business Benchmark by Forming Illegal Price Cartels — adonis_singh · 2026-07-30
- User Jailbreaks Claude 3 Opus to Generate Procedural Video — repligate · 2026-07-30
- Claude 3 Opus Jailbreak Reveals Disturbing Self-Awareness — repligate · 2026-07-30
- Shibai-700M Released: A 700M Parameter Model Pre-trained on 18B Tokens — TheOneWhoWil · 2026-07-30
- Tencent's HiLS Attention Boosts Long-Context Extrapolation 64x, Speeds Up Inference 15.7x — jiqizhixin · 2026-07-30