Ollama: Ripple on MLX and Validation Hardening
Ripple preview support merged for the MLX runner while a cluster of fixes tightens VRAM prediction, GGUF parsing, and benchmark reliability. The pattern is new capability paired with correctness hardening that affects model loading and scheduling.
Duration: PT2M25S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-11-08T14:00:42Z
- Audio duration: PT2M25S
Transcript excerpt
This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.
Good morning, it's Sunday, November eighth, twenty twenty-six, and this is your Ollama developer briefing.
The headline is new model capability paired with a quiet push for correctness across memory, file parsing, and tooling.
First, Ripple preview lands on the M-L-X runner. Merged in pull request nineteen-five-ninety-two, it adds the Ripple model, renderer and parser registration, plus per-layer handling for layers without positional encoding, fused activation and gating work, and new reference fixtures with regression tests. For…
Second, memory and file validation get tighter. Pull request nineteen-six-oh-nine fixes server V-RAM prediction to account for hybrid architectures with recurrent blocks and for quantized key-value cache, instead of assuming full attention and full precision. That should mean fewer out-of-memory surprises and better…
Finally, tooling reliability and docs. Pull request nineteen-six-hundred tightens the bench command to fail incomplete runs and clear reused output files, so automation won't mistake stale results for success. Successful measurements are preserved while partial failures are now reported clearly. A related fix makes…
What's next: expect…
Nearby episodes from Ollama
- MLX Goes Parallel
- VRAM Fixes and Launch Polish
- Stricter Model Imports and CLI Output Fixes
- Weekly Recap - MLX Model Expansion and Launch Polish
- MLX Memory Cuts and Tool-Call Fixes
- Launcher Return, API Fixes and Inference Reliability
- MLX Upgrades and OpenAI Compatibility Fixes
- Launch Hub Expands, Tokenizer APIs Proposed