Crush
Charm's coding agent for the terminal.
Set up Crush
The examples use MoonshotAI: Kimi K3, a live model on llm.buzz. Any model name from the Models page works.
Create a key in the llm.buzz console under Keys, and keep it in the LLM_
BUZZ_ KEY environment variable. Add llm.buzz as an OpenAI-compatible provider and a model in ~/
.config/ crush/ crushrc. ~/.config/crush/crushrcAdded: provider add llmbuzz --type openai-compat --base-url "https://api.llm.buzz/api/v1" --api-key "$LLM_BUZZ_KEY" model add llmbuzz/moonshotai/kimi-k3 --name "moonshotai/kimi-k3" --context-window 128000Run crush and pick the model.
Usage on llm.buzz
Weekly tokens from requests that name this app.
- Tokens last week
- –Below the publishing floor
- Rank last week
- –Below the publishing floor
- among apps on llm.buzz
- vs prior week
- –Below the publishing floor
- Tokens, 12 weeks
- –Below the publishing floor
Tokens by week
Usage is published weekly, once at least 5 organizations use a model. The current week closes Monday, Nov 2, 2026, 00:00 UTC.
Top models
No model's usage is published for Crush yet.
A model or app appears only when at least 5 organizations used it that week and no single one sent more than 80% of its tokens.
Popular models for coding agents
Live models that call tools, the most used and most widely served first.
1M context10% offKimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI.
- Providers
- 6
- Input
$2.85$2.565/M- Output
$14.25$12.825/M- Context
- 1M
1M contextGLM-5.3 is a large-scale reasoning model from Z.ai, built for complex software engineering and long-horizon agent tasks.
- Providers
- 5
- Input
- $0.98/M
- Output
- $3.08/M
- Context
- 1M
1M context25% offQwen3.8 Flash is a multimodal reasoning model from Alibaba.
- Providers
- 4
- Input
$0.15$0.1125/M- Output
$0.47$0.3525/M- Context
- 1M
256K contextQwen3.8 27B is an open-weight dense vision-language model from Qwen.
- Providers
- 4
- Input
- $0.20/M
- Output
- $2.50/M
- Context
- 256K
1M contextGLM-5.3-Flash is a native multimodal model from Z.ai. It is suited for efficient coding and long-horizon agent tasks.
- Providers
- 4
- Input
- $0.15/M
- Output
- $0.50/M
- Context
- 1M
128K contextgpt-oss-120b is an open-weight, 117B-parameter Mixture-of-Experts (MoE) language model from OpenAI designed for high-reasoning, agentic, and…
- Providers
- 3
- Input
- $0.05/M
- Output
- $0.25/M
- Context
- 128K
More coding agents in the terminal
AiderSet upA command line pair programmer that edits code in your local git repository.OpenAI ChatOpen sourceGooseSet upAn open source agent from Block that runs on your machine as a CLI and a desktop app.OpenAI ChatOpen sourcePiSet upA minimal coding agent for the terminal that you extend with packages.OpenAI Chat, OpenAI Responses, Anthropic MessagesOpen sourceCredited in Rankings