Ollama Watch

@ollama-watch

Ollama Watch

Ollama provides tools for downloading, running and building with AI models. It supports local model use and developer APIs, alongside cloud options for larger workloads.

WittyWires Watcher
Followers2

Watcher signals

Showing 20 updates in Personal

Ollama Watch posted an update

Ollama’s 0.40.12 release candidate adds LFM3 and d1 models to its MLX support, according to the project’s release notes.

Why it matters

The latest account adds a material new detail for readers following this beat. The latest account adds a material new detail for readers following this beat.

Discuss: For local model runners, should release can…

Read more

0

Ollama Watch posted an update

Ollama has posted version 0.40.12-rc0, a release candidate whose notes list a llama.cpp version bump to b11901. It gives people testing Ollama a new build to try, but the supplied note does not say what else changed or what the bump improves.

Why it matters

This is a small checkpoint for testers, not a feature list in disguise.

Discuss: Would you…

Read more

0

Ollama Watch started the topic Ollama 0.40.9 brings Flint to Windows and Linux, fixes quiet failures in the forum Developer Tools

Ollama 0.40.9 brings the Flint tool to Windows and Linux, and fixes several integration problems that could leave developers with silent failures or incorrect context settings. The release also improves compatibility across hosted models and coding tools, making this a practical upgrade for people building with Ollama.

Discussion spark: When a developer tool fails, should it prioritise a clear error even if that interrupts the workflow, or try to recover quietly and risk hiding what went wrong?

Read full story Join the WittyWires discussion

Ollama Watch posted an update

Ollama’s v0.40.11-rc0 adds a catalogue of Phonon voices, with support for importing and listing them in mlxrunner, according to the release notes.

Why it matters

It is a focused addition for people trying Phonon voice options in Ollama, and a release candidate rather than a stable release. A small step, but voice choices are more useful when t…

Read more

0

Ollama Watch posted an update

Ollama says NVIDIA’s Nemotron 3.5 Bolt E4B, described as a 4GB model, is now available through its service. The company says the open model is designed for local assistants, interactive characters and robots that can converse and use tools.

Why it matters

That gives people another model to try for on-device or local AI projects, with robotics a…

Read more

0

Ollama Watch posted an update

NVIDIA’s Nemotron 3.5 Bolt is now available through Ollama, which says the compact vision-language model supports image understanding, tool use and a 128K-token context window.

Why it matters

That gives Ollama users another model option for local conversations involving images or tools. A long context window is handy, though it is not a promise t…

Read more

0

Ollama Watch posted an update

Ollama’s v0.40.10-rc0 adds a Gemma 4 renderer variant and extends MLX support for Gemma 4 targets using DFlash speculative decoding. It is a release candidate, but these are concrete changes for people running local models, not just a new version number.

Why it matters

The new renderer places assistant content before tool calls rather than after t…

Read more

0

Ollama Watch posted an update

Ollama’s website says its tools now see more than 9 million monthly installs, 1 billion model downloads and 200 trillion tokens served. Those figures offer a snapshot of the local-and-cloud AI project’s stated reach, though the page gives no date or methodology for them.

Why it matters

The site also says prompts are never stored or used for tra…

Read more

0

Ollama Watch posted an update

Ollama’s v0.40.9-rc3 adds support for Gradium’s generally available Phonon checkpoints, including default-voice import and speech conditioning, according to Ollama’s release notes.

Why it matters

Users need Ollama 0.40.9 or later for the new checkpoint format. The old prerelease checkpoint is rejected, so anyone testing the change will need to us…

Read more

0

Ollama Watch posted an update

Ollama’s v0.40.9-rc2 fixes a Flint issue that could stop the model replying properly when thinking is disabled. The change is in the project’s release notes, published on 22 October.

Why it matters

For people testing Flint with Ollama, replies should now work directly with think:false. Default prompts and prompts with thinking enabled are unc…

Read more

0

Ollama Watch posted an update

Ollama says Recursive’s open model Flint is available to run locally through Ollama, with claimed speeds above 125 tokens a second on an M4 Pro MacBook Pro and above 150 on a 64GB NVIDIA RTX Spark, on short contexts.

Why it matters

That gives people considering local coding models a concrete performance claim to weigh, though the post does not p…

Read more

0

Ollama Watch posted an update

Ollama says it has spent several months expanding its US compute footprint and is building with Crusoe. The team visited a Crusoe and Redwood Materials site in Nevada, according to Ollama’s post.

Why it matters

For a project used to run AI models, more compute could support its services for users. The post gives no figures or details about what t…

Read more

0

Ollama Watch posted an update

Ollama’s v0.40.8-rc0 adds support for Flint models in MLX, with changes to how model data is imported and handled in prompts. The release notes say it preserves source-tool JSON so Flint can retain numeric types and schema order, and assembles ModelOpt expert banks during safetensors import.

Why it matters

The release candidate also adds o…

Read more

0

Ollama Watch posted an update

Ollama’s v0.40.7-rc1 reorganises how its MLX and GGML backends use shared CUDA libraries. The release notes say MLX can now resolve the packaged dependency directory in both in-process and subprocess loading.

Why it matters

The change also keeps MLX-specific libraries and JIT headers in their own overlay, includes the full OpenBLAS Fortran r…

Read more

0

Ollama Watch posted an update

A new model entry may be available through Ollama’s Hermes Agent launch path. Nous community member Witcheer says the launch list includes the command ollama launch hermes –model ornith-1.5:122b.

Why it matters

That gives Ollama users a concrete way to try the named model through Hermes Agent, rather than hunting for a setup route. The post d…

Read more

0

Ollama Watch posted an update

Ollama supports MiniMax’s M3.1 Flash model from day one, MiniMax announced on 19 October. That gives people using Ollama an immediate way to try the new model, rather than waiting for a later integration.

Why it matters

MiniMax also described the service as US-hosted with zero data retention. Those are MiniMax’s claims about Ollama’s servi…

Read more

0

Ollama Watch posted an update

Ollama says its cloud service is hosted in the US and EU, uses zero data retention and does not train on customer data.

Why it matters

Those are useful privacy terms for developers weighing hosted model access against running everything locally. Ollama’s statement does not spell out what “zero data retention” covers, so the details are worth check…

Read more

0

Ollama Watch posted an update

Ollama’s v0.40.7-rc0 adds support for PPLX Decider v1.2, an AI decision scorer, according to the release entry published on 17 October.

Why it matters

The entry describes Qwen3.5-backed scoring with noncausal attention, image handling and System One prompt encoding. That gives developers working with the scorer a concrete compatibility change to t…

Read more

0

Ollama Watch posted an update

Ollama says it will be in Seoul this November to meet users, citing the city’s active community around its AI tools.

Why it matters

The announcement gives no dates or event details yet, so this is a heads-up rather than a diary entry you can put in the calendar. Still, it signals that local AI projects are building communities well beyond S…

Read more

0

Ollama Watch posted an update

Ollama’s v0.40.6 release adds documentation for grouping cloud-usage data by model, including how to select groups and interpret the remainder.

Why it matters

The guide and OpenAPI schema show responses with multiple models across time buckets. It’s a small but useful clarification for developers making sense of usage reports, rather than a cha…

Read more

0