Ollama Release Notes

Muse Glimmer

Muse Glimmer is now available on all platforms. Muse Glimmer can power coding agent applications such as Claude Code, Codex, Pi and more, as well as long-running personal assistants such as OpenClaw and Hermes.

Ollama's MLX engine provides state-of-the-art performance on Apple Silicon for this model, with support for DFlash and image input as of Ollama 0.32.7.

To download and run Muse Glimmer locally:

ollama run muse-glimmer

To run Muse Glimmer with Claude Code, download Ollama and run:

ollama launch claude --model muse-glimmer

For a lighter-weight coding agent, try Pi:

ollama launch pi --model muse-glimmer

For personal assistant frameworks such as OpenClaw and Hermes, use:

ollama launch openclaw --model muse-glimmer
ollama launch hermes --model muse-glimmer

What's Changed

Add Muse Glimmer support for NVIDIA, AMD, and additional platforms

Full Changelog: v0.32.7...v0.32.8

Muse Glimmer

Note: Muse Glimmer is currently available via initial support via Ollama's MLX engine on Apple Silicon. Additional support and optimizations for Apple Silicon, NVIDIA, AMD, and other platforms will be available in the coming days.

Muse Glimmer, Meta's newest open model and the first released by Meta Superintelligence Labs, is now available on Ollama. It's a 30B multimodal model purpose-built for agent workloads that run locally.

With Ollama, you can now use Muse Glimmer to power coding agent applications such as Claude Code, Codex, Pi and more, as well as long-running personal assistants such as OpenClaw and Hermes.

Ollama's MLX engine provides state-of-the-art performance on Apple Silicon for this model, with support for DFlash and image input as of Ollama 0.32.7.

To download and run Muse Glimmer locally:

ollama run muse-glimmer:30b-mlx

To run Muse Glimmer on Apple Silicon with Claude Code, download Ollama and run:

ollama launch claude --model muse-glimmer:30b-mlx

For a lighter-weight coding agent, try Pi:

ollama launch pi --model muse-glimmer:30b-mlx

For personal assistant frameworks such as OpenClaw and Hermes, use:

ollama launch openclaw --model muse-glimmer:30b-mlx
ollama launch hermes --model muse-glimmer:30b-mlx

What's Changed

Qwen3.5 is faster on Apple GPUs: the MLX engine now uses the model's MTP head for speculative decoding automatically

/v1/chat/completions streaming now matches OpenAI's wire format: role only on the first chunk, finish_reason on its own chunk, and usage in a separate chunk with stream_options.include_usage.

Truncated OpenAI responses now report finish_reason: "length" instead of "tool_calls".

ollama run kimi-k3 now offers kimi-k3:cloud for cloud-only models that publish no default tag, instead of failing.

TUI fixes: pipe-delimited prose no longer renders as a table, Enter accepts the highlighted @ file completion, and /prompt scrolling is no longer laggy.

Experimental image generation has been temporarily removed. Continue using 0.32.5 for image generation support

Updated the MLX and llama.cpp engines.

Full Changelog: v0.32.5...v0.32.6-rc0

model: add Laguna MLX support

Add Laguna XS 2, XS 2.1, and S 2.1 support to the MLX model and create paths.

Read the source config to apply one quantization policy across dense and routed MoE layers. Keep the tied output head and router at source precision, quantize supported attention and expert projections, selectively promote sensitive expert down projections, and emit per-tensor metadata for mixed quantization blobs.

Correct dense expert loading, BF16 source-layout handling, expert global-scale shapes and dtypes, routing-score scaling, and mixed-precision expert dispatch. Gate/up and down projections select quantized or dense execution independently so promoted BF16 down projections do not force quantized gate/up weights through the dense fallback.

Optimize the forward pass with compatible gate/up fusion, sorted standard GatherMM and GatherQMM operations for larger prefills, model-local mlx.Compile closures for elementwise MoE work, and cache-backed 512-token prefill chunks. This keeps the implementation on maintained MLX operations without custom kernels.

Add focused tests for Laguna configuration variants, quantization policy and metadata, dense and routed expert loading, mixed-precision dispatch, compiled-versus-eager parity, fused projections, routing, and prefill chunking.

review comments and S 2.1 performance fixes

Address renderer/parser selection and mixed-precision expert quantization review feedback.

Keep Laguna weights resident on Metal to prevent repeated paging of its large, sparsely accessed expert buffers. Scope this policy to Laguna GPU execution.

Remove obsolete 512-token prefill chunking now that the runner's 2048-token path is faster.

review comments addressed

fix create

What's Changed

Fixed model downloads that stall before sending data.

Improved integrations: restored Claude Code Channels, fixed Anthropic thinking streams, and made Hermes Desktop respect --force-build.

Expanded GPU support with CUDA on Windows ARM64, B200 support through CUDA 12, and lower memory use on Linux CUDA/ROCm iGPUs.

Added chat, thinking, and tool calling support for Laguna 2.1 models, including a Metal inference fix.

Fixed GLM tool calls being silently dropped at the end of generation.

Updated the MLX and llama.cpp engines.

Full Changelog: v0.32.1...v0.32.3