What's Changed
- fix(core): only honor tool-call and reasoning delimiters emitted as special tokens by @EricLBuehler in #2455
- feat(models): add support for Qwen3.8-Flash-Next by @EricLBuehler in #2462
- feat(cuda): support CUDA 13.4 by @EricLBuehler in #2481
- fix(cuda): support GCC 13+ in radix top-k kernel by @wsj040303 in #2461
- fix(quant): avoid deadlock while batching ISQ output jobs by @mikemikimike in #2463
- feat(metal): make Metal Paged Attention the default backend by @EricLBuehler in #2493
- chore: v0.9.4 -> v0.9.5 by @EricLBuehler in #2498
New Contributors
- @wsj040303 made their first contribution in #2461
- @mikemikimike made their first contribution in #2463
Full Changelog: v0.9.4...v0.9.5