First-timer quantizes DeepSeek 4.1 with ExLlamaV3's EXL3, with an AI copilot
HankYeomans · x · 2026-11-19
A developer shares his first attempt at model quantization: targeting DeepSeek 4.1 using ExLlamaV3's EXL3 quantization. He admits he's inexperienced and the process may be messy, but argues the point is participating hands-on rather than just picking up one of thousands of pre-quantized models.
He also credits Ampcode as a learning partner that keeps asking follow-up questions while helping him build a "brain" and training setup — and notes it's the first AI provider he feels comfortable talking to by voice instead of typing.
More from Models
- Anthropic's web search blocks Reuters, WSJ, FT — irony noted by users — ssh4net · 2026-11-19
- Karpathy unveils nanochat miniseries v1, treating LLMs as a scaling-law model family with one compute dial — eliebakouch · 2026-11-19
- This year's most consequential interpretability work relied on open models — georgejrjrjr · 2026-11-19
- UK Turing Institute sovereign AI model roadmap targets deployable models late 2028 onwards — bclavie · 2026-11-19
- Gemini release gaps halve each cycle: 28 to 15 to 6 days per new model — davidyin44 · 2026-11-19
- Surge AI benchmarks Gemini 4 Neon: tops Tuesday@Work at 71.6, trails on SWE — echen · 2026-11-19