Tag: Quantization
-
Google Puts TPUv7 Through an Outside Agentic Benchmark
by Chris DePuy, 650 Group / October 25, 2026 Google’s first TPUv7 submission to an outside agentic-inference benchmark landed this week. The operator’s October 14, 2026 announcement reports FP8 results served through vLLM at 1.67x better performance per dollar than NVIDIA’s GB300 NVL72 at FP4 the operator’s announcement thread. The submission is the first submitted…
-
New TensorFold LLM Model Serving Software Dramatically Improves Decode Speed on Macs and Sparks
by Chris DePuy, 650 Group / September 27, 2026 Two repositories shipped from one maintainer this week and made the same promise in two different places. On September 25 and 26, a developer publishing under the name Ash Hart released TensorFold, a local inference engine for Apple Silicon and NVIDIA GPUs, and Imprint, a tool…
-
NVIDIA’s Official NVFP4 Checkpoints for DeepSeek-V4.1-Flash and GLM-5.3-Flash: Four DGX Sparks Read the Receipts
by Chris DePuy / September 22, 2026 NVIDIA: what the conversion actually changes The DeepSeek checkpoint converts only the ordinary routed experts — projections w1, w2 and w3 across 384 experts in 40 layers — from the source MXFP4 format to NVFP4 weights and activations at group size 16, per the model card; attention, shared…
-
GLM-5.3-Flash on a Single DGX Spark: Three Community Quantizations Fit Where the Vendor’s Does Not
by Chris DePuy / September 13, 2026 On September 9, NVIDIA published an official NVFP4 quantization of Z.ai’s GLM-5.3-Flash. Its Model Optimizer recipe leaves the key-value cache at FP8 and compresses only the linear layers inside the transformer blocks — the sparse-MoE shared experts and the dense MLP — from 16 bits down to 4,…
-
DFlash2 Beats DSpark for Qwen3.8-27B on One DGX Spark — and Jumps to Apple Silicon
by Chris DePuy / August 20, 2026 A week ago the Qwen3.8-27B decode race on a single NVIDIA DGX Spark was a contest between DSpark and MTP. This week the leading recipes on the same hardware have switched to a different engine: DFlash2, the block-diffusion speculative decoder that first shipped with DeepSeek, is now beating…
-
GLM-5.3-Flash Day Two: DFlash 2 Attacks the Decode Path While 1M-Context KV Moves Onto Two DGX Sparks
by Chris DePuy / August 28, 2026 A day after Z.ai opened weights on GLM-5.3-Flash, the desk-side community pushed the model past what I thought was its practical limit on small hardware. Two bottlenecks — the decode path and the KV cache — were attacked in parallel, and both moved in the same 24 hours.…
-
1T Inkling Model at 1-bit is 86% smaller
AI-generated concept image of model compression via 1-bit quantization. Thinking Machines Lab’s Inkling — a 975B-parameter (41B active) multimodal with 1M context, vision, and audio (HF model card) — launched day one with a dynamic 1-bit GGUF courtesy of Unsloth. The Compression Format Size Vs. Full BF16 (full) 1.9 TB — NVFP4 ~490 GB 74%…