Models

Gluon’s built-in catalog: facts about each model, not configuration. The Id is what the intake agent proposes and --launch --model takes; the connection columns are the id each connection’s API or plan uses (— : not served there). Which of these the intake agent may propose is decided per machine by a check (gluon doctor); how it decides is in Connections.

Claude Code

Anthropic’s own harness. Strongest general coding agent; best for multi-file features, debugging and refactors.

• Binary: claude (Anthropic)
• Connections: Claude plan (subscription), Anthropic, Amazon Bedrock, OpenRouter
• Modes besides build: explore: supported; plan: supported

Model	Id	Claude plan	Anthropic	Amazon Bedrock	OpenRouter	Efforts	Default effort	Notes
Haiku 5.5	haiku	haiku	claude-haiku-5-5	global.anthropic.claude-haiku-5-5	anthropic/claude-haiku-5.5	low, medium, high, xhigh, max	medium	fast and cheap; small, well-specified edits
Sonnet 5.5	sonnet	sonnet	claude-sonnet-5-5	global.anthropic.claude-sonnet-5-5	anthropic/claude-sonnet-5.5	low, medium, high, xhigh, max	high	strong and mid-priced; default for most everyday tasks
Opus 5.5	opus	opus	claude-opus-5-5	global.anthropic.claude-opus-5-5	anthropic/claude-opus-5.5	low, medium, high, xhigh, max	medium	most capable, expensive; hard debugging, design, large changes
Fable 5.1	fable	fable	claude-fable-5-1	global.anthropic.claude-fable-5-1	anthropic/claude-fable-5.1	low, medium, high, xhigh, max	high	frontier model, most expensive; the hardest, longest tasks

Install (official methods, the vendor’s recommended first; Gluon shows the command and runs it only on an explicit “Run it”):

Where	Command	Downloads from	Note
Linux, macOS, WSL	curl -fsSL https://claude.ai/install.sh | bash	claude.ai
Windows	irm https://claude.ai/install.ps1 | iex	claude.ai

Codex

OpenAI’s harness. Strong at well-specified implementation work and careful, test-driven changes.

• Binary: codex (OpenAI)
• Connections: ChatGPT plan (subscription), OpenAI, Amazon Bedrock, OpenRouter
• Modes besides build: explore: supported; plan: typed as /plan once the agent is up

Model	Id	ChatGPT plan	OpenAI	Amazon Bedrock	OpenRouter	Efforts	Default effort	Notes
GPT-6 Luna	gpt-6-luna	gpt-6-luna	gpt-6-luna	global.openai.gpt-6-luna	openai/gpt-6-luna	low, medium, high, xhigh	high	fast and cheap; small, well-specified edits
GPT-6.1 Sol	gpt-6.1-sol	gpt-6.1-sol	gpt-6.1-sol	global.openai.gpt-6.1-sol	openai/gpt-6.1-sol	low, medium, high, xhigh	medium	strong and mid-priced; everyday implementation work
GPT-6 Astra	gpt-6-astra	gpt-6-astra	gpt-6-astra	global.openai.gpt-6-astra	openai/gpt-6-astra	low, medium, high, xhigh	medium	most capable, expensive; hard, ambiguous work

Install (official methods, the vendor’s recommended first; Gluon shows the command and runs it only on an explicit “Run it”):

Where	Command	Downloads from	Note
Linux, macOS, WSL	curl -fsSL https://chatgpt.com/codex/install.sh | sh	chatgpt.com
Linux, macOS, WSL	npm install -g @openai/codex	registry.npmjs.org
Windows	irm https://chatgpt.com/codex/install.ps1 | iex	chatgpt.com
Windows	npm install -g @openai/codex	registry.npmjs.org

Antigravity

Google’s harness on Gemini. Fast and cheap; good for well-scoped edits and quick fixes.

• Binary: agy (Google)
• Connections: Google account (subscription), Gemini API
• Modes besides build: explore: plan mode; plan: supported

Model	Id	Google account	Gemini API	Efforts	Default effort	Notes
Gemini 3.8 Flash	gemini-3.8-flash	gemini-3.8-flash	gemini-3.8-flash	low, medium, high	high	fast Gemini; good quality at a low price

Install (official methods, the vendor’s recommended first; Gluon shows the command and runs it only on an explicit “Run it”):

Where	Command	Downloads from	Note
Linux, macOS, WSL	curl -fsSL https://antigravity.google/cli/install.sh | bash	antigravity.google
Windows	irm https://antigravity.google/cli/install.ps1 | iex	antigravity.google

Grok Build

xAI’s harness on Grok. Quick, capable generalist; good value for everyday coding tasks.

• Binary: grok (xAI)
• Connections: SuperGrok / X account (subscription), xAI
• Modes besides build: explore: supported; plan: typed as /plan once the agent is up

Model	Id	SuperGrok / X account	xAI	Efforts	Default effort	Notes
Grok 4.7	grok-4.7	grok-4.7	grok-4.7	low, medium, high, xhigh	high	xAI’s newest; capable generalist

Install (official methods, the vendor’s recommended first; Gluon shows the command and runs it only on an explicit “Run it”):

Where	Command	Downloads from	Note
Linux, macOS, WSL	curl -fsSL https://x.ai/cli/install.sh | bash	x.ai
Windows	irm https://x.ai/cli/install.ps1 | iex	x.ai

OpenCode

Open-source harness on OpenRouter or the OpenCode Go plan. Good for well-scoped tasks when an open or cheaper model is enough.

• Binary: opencode (OpenCode)
• Connections: OpenRouter, OpenCode Go plan
• Modes besides build: explore: supported; plan: supported

Model	Id	OpenRouter	OpenCode Go plan	Efforts	Default effort	Notes
DeepSeek Flash	deepseek-flash	deepseek/deepseek-v4.1-flash	deepseek-v4.1-flash	low, high, max	max	very cheap and fast; small, mechanical edits (follows the latest DeepSeek Flash: V4.1 today); follows the latest release (deepseek-v4.1-flash today)
Muse Spark 1.3	muse-spark-1.3	meta/muse-spark-1.3	—	low, medium, high, xhigh, max	max	Meta’s model; capable generalist
Muse Spark 1.3 Contributor	muse-spark-1.3-contributor	meta/muse-spark-1.3-contributor	muse-spark-1.3-contributor	low, medium, high, xhigh	xhigh	Meta’s coding-tuned variant, very cheap; Meta receives the session’s code and prompts; opt-in: allow_muse_contributor in routing.yaml; Meta receives the session’s code and prompts

Install (official methods, the vendor’s recommended first; Gluon shows the command and runs it only on an explicit “Run it”):

Where	Command	Downloads from	Note
Linux, macOS, WSL	curl -fsSL https://opencode.ai/v2/install | bash	opencode.ai
Linux, macOS, WSL	npm install -g @opencode/cli	registry.npmjs.org
Windows	npm install -g @opencode/cli	registry.npmjs.org

Kimi Code

Moonshot AI’s harness on Kimi K3 and K2.7 Code, on a Moonshot or OpenRouter API key, or the Kimi Code plan. Good for long agentic coding tasks at a modest price.

• Binary: kimi (Moonshot AI)
• Connections: Kimi Code plan (subscription), Moonshot AI, OpenRouter
• Modes besides build: explore: unavailable (its interactive mode ignores an agent file, so no flag can take its writing and shell tools away; use plan mode, or build); plan: supported

Model	Id	Kimi Code plan	Moonshot AI	OpenRouter	Efforts	Default effort	Notes
Kimi K2.7 Code	kimi-k2.7-code	—	kimi-k2.7-code	moonshotai/kimi-k2.7-code	none	—	Moonshot’s coding model; cheap, good for well-scoped coding tasks (not on the plan)
Kimi K3	kimi-k3	k3	kimi-k3	moonshotai/kimi-k3	low, high, max (Kimi Code plan only)	high	Moonshot’s flagship; long agentic coding tasks, 1M context

Install (official methods, the vendor’s recommended first; Gluon shows the command and runs it only on an explicit “Run it”):

Where	Command	Downloads from	Note
Linux, macOS, WSL	curl -fsSL https://code.kimi.com/kimi-code/install.sh | bash	code.kimi.com	Linux needs glibc: the installer refuses Alpine (musl). It adds kimi to your shell profile’s PATH and renames an old Python kimi-cli on PATH to kimi-legacy.
Windows	irm https://code.kimi.com/kimi-code/install.ps1 | iex	code.kimi.com	Needs Git for Windows (kimi’s shell). It renames an old Python kimi-cli on PATH to kimi-legacy.