Which agents work with it?

Any agent that reads skills from a directory. Claude Code, Codex and Gemini CLI read it natively; Cursor and VS Code connect over MCP.

The voice runs on your own machine, so there is nothing to meter. The first run downloads it, about 340 MB.

Yes. The skill is MIT-licensed and the authoring format is public.

Create

Ollama v0.40.15 runs Xiaomi's MiMo V2.6 and AutoTrust models that answer yes or no

13 hours ago

@ollama Subscribe

Ollama v0.40.15 runs Xiaomi's MiMo V2.6, a 9B model for code and agents, and AutoTrust decision models that answer a question with a choice, a yes or no probability, or a rating (Apple silicon only for now). ollama launch sets up Kimi Code CLI, and recurrent and vision GGUF models serve parallel requests. https://github.com/ollama/ollama/releases/tag/v0.40.15 https://github.com/ollama/ollama/compare/v0.40.14...v0.40.15

Ask about this presentation

Chapters

0:00Ollama now runs Xiaomi’s MiMo V2.6, a 9B model for code and agents
0:13AutoTrust models answer your question with a choice, a probability or a rating
0:28Ollama launch now sets up Kimi Code CLI, the successor to kimi-cli
0:41Recurrent and vision GGUF models now serve more than one request at a time
0:55Long-context Qwen works on MLX, Pi lists your models, and plain ollama opens the launcher
1:10Pull MiMo V2.6 if you want a small local model for coding agents

Show transcript

Ollama now runs MiMo V2.6, Xiaomi's nine billion parameter model built for coding and agent work. One command pulls it and starts a chat: ollama run MiMo V2.6.
Slide text
Ollama
NOV 3, 2026

Ollama now runs Xiaomi’s MiMo V2.6, a 9B model for code and agents

$ ollama run mimo-v2.6

Ollama release notes, Nov 3 · PR #19400
AutoTrust decision models read text and answer your question with a choice, a yes or no probability, or a rating. Ask whether a complaint is a billing issue, and it tells you. For now they run on Apple silicon only.
Slide text
Ollama
NOV 3, 2026

$ ollama run autotrust-jev "A customer says their card was charged twice." --ask "Is this a billing issue?"

Apple silicon only in this release. The 35b and 27b-vl sizes also read images.

Ollama release notes, Nov 3 · PR #19399
Ollama launch Kimi now sets up Kimi Code CLI, the replacement for the archived Kimi CLI. If you still have the old one, Ollama offers to move you over, and saying no keeps it working.
Slide text
Ollama
NOV 3, 2026

$ ollama launch kimi

Still on kimi-cli? Ollama offers to install Kimi Code. Say no and the old CLI keeps working.

Ollama release notes, Nov 3 · PR #19506
Recurrent and vision models in GGUF format can now answer several requests at once. They were held to one, because Ollama's older engine kept a single sequence per cache. Llama server has room for more.
Slide text
Ollama
NOV 3, 2026

Before: held to one request at a time

Now: parallel requests through llama-server

Ollama release notes, Nov 3 · PR #19545
Qwen 3 and Qwen 3.5 models that use YaRN for long context now get it in Ollama's MLX engine. Pi lists your Ollama models by itself, and Claude Desktop shows their real names. And plain ollama opens the launcher again.
Slide text
Ollama
NOV 3, 2026

Qwen3 and Qwen3.5 get their YaRN long-context scaling on MLX

Pi fills in its model list, and Claude Desktop shows real model names

ollama with no arguments opens the welcome launcher

Ollama release notes, Nov 3 · PRs #19231, #19460, #19533, #19478
From this release, MiMo V2.6 is the one I would pull, because it is the one model here built for coding agents. Ollama's release notes list every change.
Slide text

Ollama
NOV 3, 2026

Ollama release notesgithub.com/ollama/ollama/releases

github.com/ollama/ollama/releases

Up next