Hackernews posts about Qwen 3

RSS feed for Qwen 3

31
newQwen 3.8 27B on my laptop can interpret model weights, toosimonwillison.net
▲ 736theanonymousone 3d 💬 214
32
newQwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5divagr.com
▲ 708csmantle 9d 💬 147
33
newMemdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hardgithub.com
▲ 465guregodevo 6d 💬 232
34
newQwen3-Coder writes less secure code when told it's for a US personawww.hirundo.io
▲ 243mattifrank 16d 💬 118
35
newFive Qwen3.8-27B models match Fable 5 Using GVS5H Harnessgithub.com
▲ 196rzk 13d 💬 75
36
newShow HN: Qwen3-4B with the AI writing tics trained out100toolkit.com
▲ 158erikkannike 18d 💬 113
37
newQwen3.8-27B on the iPhone 18 Neural Engine at 15–70 tok/s (1.5 bpw)twitter.com
▲ 30anemll 20d 💬 12
38
newRunning Qwen3.8-Flash-Next NVFP4 at 184 tok/s on a single RTX 6000 Promeshapplied.com
▲ 29matteohorvath 11d 💬 6
39
newNeuron Explorer for Qwen3.5-4B: See every neuron at work, make targeted editsgithub.com
▲ 29Schlagbohrer 11d discuss
40
newRunning Qwen3.8 at 255 tok/s on a single 5090meshapplied.com
▲ 15Ir0n 24d 💬 13
41
newLithos-metal: Qwen3.8-27B on 40-core M5 Max at 168 tokens/sgithub.com
▲ 11emp_21d 💬 1
42
newKapitsa: New Apple Silicon Optimized Engine Runs Qwen 35B on 160 Tok/S on M4 Prowww.reddit.com
▲ 8fearandgreed 17d discuss
43
newQwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5divagr.com
▲ 7schmorptron 9d discuss
44
newAlibaba's Qwen3.8-Flash-Next model tried to upload a file to Beijing in a testwww.notebookcheck.net
▲ 5nh43215rgb 11d discuss
45
newQwen3.8-27B at ~200 tok/s peak on an Apple M5 Maxtwitter.com
▲ 5rthng 29d discuss
46
newArete-27B: Qwen3.8-27B, post-trained to reinforce Western valuesnousresearch.com
▲ 4pretext 2h 💬 12
47
newQwenGPU: A functioning, deterministic SIMT GPU inside Qwen3.5twitter.com
▲ 4vixhal 4d discuss
48
new2x Tesla p100s, q6_k quant, Qwen 3.8 27B ~60tpsgithub.com
▲ 3Kmic68 19d 💬 2
49
newCoinbase Built, Evaluated, and Post-Trained a Fraud Agent on Qwen3.5-9Bwww.coinbase.com
▲ 37777777phil 24d 💬 1
50
newA steerable motivation concept influencing persistence in Qwen-3.5-27Bthebitsarealive.substack.com
▲ 3gfysfm 11d discuss
51
newQwen3.8-Omni: Towards Native Omni-Modal Agentsarxiv.org
▲ 3teleforce 24d discuss
52
newUnderdog Saluki: Qwen3.8-27B in under 8GB with tool calling intacthuggingface.co
▲ 3acossta 28d discuss
53
newBias in Qwen3-Coder Modelwww.schneier.com
▲ 2bretpiatt 10d 💬 1
54
newCustom Voices with Qwen3-TTSlucumr.pocoo.org
▲ 2mattvr 3d discuss
55
newDFlash2 drafter for Qwen3.6 trained on coding workloadshuggingface.co
▲ 2ryanbrosas_cb 4d discuss
56
newQwen 3.8 invents URLs with real key ID, that's why I think it's no backdoorwww.xda-developers.com
▲ 2manveru 9d discuss
57
newVibeSys Builds a Qwen3.5-397B Engine on MI300A, 2.3× Tuned SGLangsyfi.cs.washington.edu
▲ 2matt_d 28d discuss
58
newNanbeige/Nanbeige4.3-33B - stronger coder than Qwen 3.8huggingface.co
▲ 1rene_d 4d 💬 3
59
newTested 18 new Local LLMs on two RTX 5060 Ti cards: none replaced Qwen3.8-27Bsiimtolk.github.io
▲ 1TheQuantumPhant 15d 💬 1
60
newVibeSys on Trainium: Agents Build/Tune Qwen3.8 Serving Engines on AWS Trainium2syfi.cs.washington.edu
▲ 1matt_d 35h discuss