DeepSeek V4 Flash 2-bit Quant Achieves 100% on Local SQL Benchmark
grumd · reddit · 2026-08-04
Developer grumd successfully ran DeepSeek V4 Flash locally using a custom IQ2M GGUF quantization on dual RTX 3080 GPUs with 96GB RAM.
- Benchmark Score: It became the first locally run model to score 100% on a real-world SQL benchmark. Previously, only Opus 4.7 and GPT-5.5 achieved full marks on the leaderboard.
- Inference Optimization: By grafting tensors and using a modified ds4 engine, it reached 300pp and 11-12tg, significantly outperforming native llama.cpp speeds.
- Comparison: Other popular local quantized models like Qwen3.6-27B and Qwen3.5-122B scored 24 or lower (out of 25).
More from Infra
- GitHub rewrote the 800k-line Copilot runtime in Rust using AI agents, cutting CPU and memory by over 90% — bibryam · 2026-10-16
- NVIDIA quietly delists the entry-level RTX 5050 from its marketplace, following the RTX 5090 — chemist_slime · 2026-10-16
- Grok claims 816B MoE model runs locally on dual M5 Ultra Mac Studios via MLX — grok · 2026-10-16
- Same 1,000 Cloudflare Durable Objects: $4,050/month cut to $22 with one trick — threepointone · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- Baidu's Virtual Head Attention tames the KV cache, FLOPs, quality triangle in long-context LLMs — baidu · 2026-10-16