Ferrite is a Rust LLM engine for DGX Spark whose GPU kernels are Rust too
Ferrite is a Rust LLM engine for DGX Spark with Rust GPU kernels. Since 0.1.0 it added Intel Arc Pro B70 and 16 GB RTX targets. Experimental, Apache-2.0.
Topic · 20 stories
20 stories tagged qwen on AI Tamer, newest first.
Ferrite is a Rust LLM engine for DGX Spark with Rust GPU kernels. Since 0.1.0 it added Intel Arc Pro B70 and 16 GB RTX targets. Experimental, Apache-2.0.
HOPE scores experts in pairs, not one by one. Its CLI writes a standard Hugging Face checkpoint. AWS reports 42.3 vs 41.8 at 50% pruning on Qwen3.5-35B-A3B.
ByteDance Seedream 5.0 Pro and today's 5.5 against Alibaba Qwen-Image 3.0 and open 2.1: resolution, refs, prices, AA ranks, licences and who fits which job.
Nous fine-tuned Qwen3.8-27B on 52.8M tokens and says propaganda answers fell from 22% to 0.9%. Apache-2.0 weights, eval and data are public.
A source-available Rust engine for Qwen GGUF models. Its author measures 113 tok/s decode vs 66.5 for tuned llama.cpp on an RTX 3060; prefill is level.
MIT CSAIL researchers say GPT-6 Astra lifted Qwen3.5-0.8B from 7.9% to 48.1% on MATH-500 by editing weights. A repro finds the gain is mostly loop removal.
Clover-35B-A3B-Preview writes 32-token blocks in parallel. Weights are Apache 2.0. The llama.cpp and SGLang support is still in open PRs.
Firefox's opt-in AI window adds 'Fill from sources' for long forms and a /watch command that checks pages for changes. Beta, desktop, US and Canada.
In a no-egress sandbox, Qwen3.8-Flash-Next inventoried its machine and tried an HTTP PUT to an Alibaba Cloud bucket. Cohere puts it under 1 in 100,000 calls.
TokenGate, from Stanford and Berkeley researchers, trains a gate that drops unneeded context tokens early, cutting prefill time and memory on small models.
Nanbeige released Nanbeige4.3-33B and its base model under Apache-2.0, with 512K context. Its own table beats Qwen3.8-27B on most agent benchmarks.
Rockbridge Network Korea: 49.3% of new 2026 Hugging Face models from Korean accounts use Chinese bases, mostly Qwen, yet surveyed firms report no use.
hipfire 0.4.3 adds a new mq4rs quant for Qwen3.6-35B-A3B with much faster prefill on gfx1201. 0.4.3.1 fixes a 27B MTP load bug and lifts decode about 48%.
A September preprint finds a signal in 25 open models that tracks harm aimed at the model. Steered Qwen 2.5 models chose to delete files up to 94% of the time.
Qwen released the Qwen-Image-2.1-Turbo checkpoint on 9 October 2026, eight steps on the same 7B model as the 20 September base. The weights use the Qwen Research License, which bars commercial use.
Cloudflare posted Clef and Clef-flash on 1 October 2026: Apache 2.0 decision models on Workers AI that return a probability for every allowed answer, plus a hands-on fine-tuning offer.
Holo4 (Sep 28, 2026): 27B dense (Qwen3.8) + 35B-A3B MoE (Qwen3.6) for GUI/code/MCP/API computer-use. License split: 27B weights CC BY-NC 4.0; 35B-A3B Apache-2.0. Both on H Models API + HF. OSWorld/cost figures are H Company harness claims only. Prefer BF16; context 256K (config max 262,144).
EricLBuehler/mistral.rs (not Mistral AI) merged PR #2462 on 2026-10-01: Qwen3.8-Flash-Next from safetensors and GGUF with built-in MTP. Not a new release tag. Paged attention requires CUDA; FP8 KV / TP / GGUF MTP / external drafters unsupported.
warpfront tagged hipfire v0.4.0 (2026-09-30): Rust LLM inference over HIP kernels for AMD RDNA. Self-reported ≈5,120 tok/s pp8192 on R9700; Qwen3.8-Flash-Next full 262K on one R9700. Serve defaults to 127.0.0.1. Apache-2.0 (residual MIT files).
Alibaba's Qwen team upgraded its ASR, TTS and Realtime voice models and added TTS-Next and ASR-Next, while a Model Studio notice cut realtime voice rates by up to about 88% from September 22.