Strata 0.2.19 runs Inkling 2 Mini locally: 11,306 tokens/s prefill on RTX 5090
CurieuxExplorer · x · 2026-11-23
Strata 0.2.19 adds native support for Thinking Machines Lab's Inkling 2 Mini, a 26.4B-parameter looped MoE model (3.4B active per token, 1M context, text+image input, OpenAI-compatible tool calling), running locally on Windows, Linux, Apple Silicon, Intel Arc, and DGX Spark.
- Benchmarks: 11,306 tokens/s prefill and 4,087 tokens/s output on RTX 5090
- Three quantization packs: SQ-10 (10.47 GB, hybrid from 4 GiB GPUs), SQ-15 (14.35 GB, full VRAM on 16 GiB cards, 96.68% top-1 agreement), SQ-20 (19.35 GB, bit-exact lossless SF4+MX8 for 24 GiB+)
- Qwen3.8-Flash-Next gets a 40% performance boost across all supported hardware
More from Infra
- Lenovo's enterprise AI infra revenue jumps 200%, outpacing cloud giants, $80B pipeline — shashib · 2026-11-22
- Wiretok: open-source local monitor for Antigravity compaction headroom and API spend — phosphorhouse · 2026-11-22
- Running an unquantized 61GB BF16 30B model on a 12GB RTX 4070 SUPER at 0.8 tok/s — Ill-Organization-38 · 2026-11-22
- Musk details Tesla's in-car ~1B diffusion world model running at 36 fps in production — flngr · 2026-11-22
- Nvidia plunges 14% after Reuters reports SEC probe into its accounting practices — mark_k · 2026-11-22
- ING fails to sell $500M data center loan bundle, forced to sell loans one by one — churchkey · 2026-11-22