Model Releases
v0.32.10
What's Changed Models that don't set a repeat_penalty now default to 1.0 (off) instead of 1.1, matching other engines and speeding up speculative decoding; set a per-model parameter if an older model
What's Changed Models that don't set a repeat_penalty now default to 1.0 (off) instead of 1.1, matching other engines and speeding up speculative decoding; set a per-model parameter if an older model repeats itself. Faster prefill on NVFP4 MLX models with a global scale, about 7–8% on Qwen3.6 and Muse Glimmer. Fixed blob verification being skipped when an OCI manifest's config and layer share a digest. New Contributors @vigneshakaviki made their first contribution in #15504 Full Changelog: v0.32.8...v0.32.10-rc1
Related
- AdaSD: Adaptive Speculative Decoding for Efficient Language Model Inference
- TokenTiming: A Dynamic Alignment Method for Universal Speculative Decoding Model Pairs
- ConFu: Contemplate the Future for Better Speculative Sampling
- Cross-Family Speculative Decoding for Polish Language Models on Apple
Silicon: An Empirical Evaluation of Bielik11B with UAG-Extended MLX-LM - Component-Aware Self-Speculative Decoding in Hybrid Language Models
Source: Ollama Releases | 2026-08-12