Ollama: Multimodal Embeddings and Cloud Convergence
Ollama added multimodal embeddings and tightened local-cloud integration, while landing reliability fixes for Windows, OpenAI compatibility, and model listing. The pattern is expansion of the embed API paired with smoothing the cloud onboarding path.
Duration: PT2M22S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-10-07T13:01:21Z
- Audio duration: PT2M22S
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 October 7th, 2026.
The lead signal is twofold: embeddings are going multimodal, and the local server is becoming the front door for cloud accounts.
First, multimodal embeddings. PR 18820 implements a new embedding model on the M L X runner, with a 24-layer bidirectional text encoder plus shared vision and audio towers, mean pooling and normalized output. Practically, the embed API now accepts per-item media in input dictionaries, so developers can embed text…
Second, local meets cloud. PR 18829 proxies the cloud usage and balance APIs through the local server using the signed-in account, with added docs and tests for signing and disabled cloud access. That means local tooling can check quota without a separate cloud client. At the same time, PR 18826 removes the account…
Finally, a correctness sweep. PR 18777 fixes every Clef Flash request failing on Windows with non-finite logits. The cause was a 32-bit seek truncation past two gigabytes, so large weight files loaded the wrong bytes. The fix uses an explicit 64-bit seek. Related compatibility work includes support for slashes in…
What's next: validate multimodal embedding quality against text-only baselines, and…
Nearby episodes from Ollama
- MLX Performance Push and Tool-Call Reliability
- Tokenizer Parity and RC-Aware Updates
- Weekly Recap - System One Decisions Take Center Stage
- MLX Tokenizer Fixes and Thinking Model Reliability
- Tool Call Ordering Fixes and Decision Models
- Decision Models Take Shape
- Pointer-Head Scoring and Reliability Fixes
- System One Moves to Explicit Capabilities