AI Briefing#qwen
The latest AI and developer news about #qwen, with the original source and a short summary.
The latest AI and developer news about #qwen, with the original source and a short summary.
On a single RTX 3090, Qwen3.6-27B achieved a 218K context and stable tool calling.
On an RTX 5080 16GB, Qwen3.6-35B-A3B hit 30 t/s at 128K and 89 t/s for new tokens.
In an 18GB M3 Pro bench, Gemma4:e4b ranked first with 72%, while Qwen3.5:9b scored 68%.
Qwen has released Qwen-Scope, an SAE-based interpretability tool built on Qwen3 and Qwen3.5.
Re-measured the quality and long-context speed of Qwen 3.6-35B-A3B KV cache quantization on M5 Max.
IK_LLAMA has added Qwen3.5 MTP support, and MTP-preserved GGUFs are required.
Native NVFP4 in llama.cpp b8967 boosted prefill speed by up to 68%, but generation speed stayed nearly the same.
Qwen accelerated linear attention by up to 3x with TileLang-based FlashQLA.
On an M2 MacBook Pro with 32GB RAM, Qwen 3.6 27B's processing speed plummeted to 4t/s at a 52,000-token context.
Qwen Code CLI removed its built-in web_search tool and switched to MCP.
Poolside released Laguna XS.2, a 33B A3B MoE model, under Apache 2.
Running Qwen 35B-A3B always-on on a 16GB Mac M4 hit an SSD bottleneck before a RAM one.
Qwen 3.6 27B was remodeled with Opus CoT and Hermes tool-calling SFT and released in FP8.
FlashQLA optimizes GDN Chunked Prefill with CP and fused kernels, achieving up to 3x speedup over FLA Triton.
On Terminal-Bench 2.0, 27B-32B open-weight coding models recorded up to 38.2%.
Qwen3.6 27B and 35B-A3B were compared across multiple engines on 2x 5060 Ti 16GB.
Qwen 3.6 Plus and Gemini were compared in a soccer video analysis benchmark.
Applying ngram-mod to Qwen3.6 27B on the R9700 showed speed improvements, though with some variability.
Qwen3.6-27B was converted into a mixed 3-bit/5-bit quantization for MLX.
A decensored version of Qwen3.6 35B A3B using Heretic has been released.
Qwen3.6-27B was run up to a 100K context on an A5000 16GB using IQ4_XS.
On an M5 Pro 64GB, Qwen 3.6 35B A3B was faster and more accurate than the 27B.
Qwen3.6-27B ran at 85–85–100 t/s on a 24GB RTX 5090 Laptop GPU.