Providers

Providers#

We support LLMs from several providers, including OpenAI, Anthropic, OpenRouter, Requesty, Deepseek, Azure, and any OpenAI-compatible server (e.g. ollama, llama-cpp-python).

You can also bring your own subscription instead of an API key: a ChatGPT Plus/Pro plan via OpenAI Subscription or a SuperGrok plan via Grok Subscription.

Important

A provider or subscription backend is not an agent harness. This page configures model access inside the gptme runtime. For example, openai-subscription uses ChatGPT subscription access from gptme; it does not run the Codex CLI harness. Likewise, grok-subscription does not run the Grok Build harness. See 🔁 Runtime Portability for the workspace / harness / model / access distinction.

This page is an overview of how to get model access. To decide which model to use, see Models. For details, see:

Selecting a provider and model#

To select a provider and model, run gptme with the -m/--model flag set to <provider>/<model>, for example:

gptme "hello" -m openai/gpt-5.6-sol
gptme "hello" -m openai-subscription/gpt-6-astra  # uses your ChatGPT Plus/Pro subscription
gptme "hello" -m anthropic  # will use provider default
gptme "hello" -m openrouter/x-ai/grok-4
gptme "hello" -m openrouter/deepseek/deepseek-v4-flash@together  # pin to Together subprovider
gptme "hello" -m deepseek/deepseek-v4-flash
gptme "hello" -m xai/grok-4
gptme "hello" -m grok-subscription/grok-4.6  # uses your SuperGrok subscription
gptme "hello" -m gemini/gemini-2.5-flash
gptme "hello" -m groq/llama-3.3-70b-versatile
gptme "hello" -m gptme/claude-sonnet-4-6  # use the gptme managed service as router
gptme "hello" -m local/llama3.2:1b  # uses a local OpenAI-compatible server (e.g. ollama)
gptme "hello" -m custom/model  # use a custom provider plugin

You can list the models known to gptme using gptme '/models' - '/exit'.

Which tool format a model performs best with also varies by provider and model — see Tool Formats for how to choose one.

Supported providers#

Built-in providers, the model prefix to use, and how each one authenticates. Each links to its setup details on Supported Providers.

Provider

Model prefix

Authentication

OpenAI Platform

openai/

OPENAI_API_KEY

Anthropic

anthropic/

ANTHROPIC_API_KEY

OpenRouter

openrouter/

OPENROUTER_API_KEY, or browser sign-in with /account setup openrouter

Requesty

requesty/

REQUESTY_API_KEY

Groq

groq/

GROQ_API_KEY

Google Gemini

gemini/

GEMINI_API_KEY

xAI

xai/

XAI_API_KEY

DeepSeek

deepseek/

DEEPSEEK_API_KEY

Moonshot

moonshot/

MOONSHOT_API_KEY

NVIDIA

nvidia/

NVIDIA_API_KEY

Azure OpenAI

azure/

AZURE_OPENAI_API_KEY and AZURE_OPENAI_ENDPOINT

OpenAI Subscription

openai-subscription/

ChatGPT Plus/Pro sign-in with gptme-auth openai-subscription

Grok Subscription

grok-subscription/

SuperGrok sign-in with gptme-auth grok-subscription (or an existing grok login)

gptme Managed Service

gptme/

gptme-auth login, or GPTME_CLOUD_API_KEY

Local server

local/

OPENAI_BASE_URL pointing at an OpenAI-compatible server

Default models#

When you pass only a provider name (for example gptme -m anthropic), gptme uses that provider’s default model. This table is generated at docs-build time from the installed gptme, so it always reflects the current release. The summary model is the cheaper model used for conversation titles and summaries.

$ gptme-util models recommended
+---------------------+--------------------------------------------+--------------------------------------------+
| Provider            | Recommended model                          | Summary model                              |
+=====================+============================================+============================================+
| openai              | openai/gpt-5.6-sol                         | openai/gpt-5-mini                          |
+---------------------+--------------------------------------------+--------------------------------------------+
| openai-subscription | openai-subscription/gpt-6-astra            |                                            |
+---------------------+--------------------------------------------+--------------------------------------------+
| anthropic           | anthropic/claude-sonnet-4-6                | anthropic/claude-haiku-4-5                 |
+---------------------+--------------------------------------------+--------------------------------------------+
| openrouter          | openrouter/deepseek/deepseek-v4-flash-0731 | openrouter/deepseek/deepseek-v4-flash-0731 |
+---------------------+--------------------------------------------+--------------------------------------------+
| gptme               | gptme/claude-sonnet-4-6                    |                                            |
+---------------------+--------------------------------------------+--------------------------------------------+
| gemini              | gemini/gemini-3.1-pro-preview              | gemini/gemini-2.5-flash                    |
+---------------------+--------------------------------------------+--------------------------------------------+
| groq                | groq/llama-3.3-70b-versatile               |                                            |
+---------------------+--------------------------------------------+--------------------------------------------+
| xai                 | xai/grok-4.6                               | xai/grok-4-1-fast                          |
+---------------------+--------------------------------------------+--------------------------------------------+
| grok-subscription   | grok-subscription/grok-4.6                 |                                            |
+---------------------+--------------------------------------------+--------------------------------------------+
| deepseek            | deepseek/deepseek-v4-flash                 | deepseek/deepseek-v4-flash                 |
+---------------------+--------------------------------------------+--------------------------------------------+

Subscriptions#

Several frontier models are reachable through a consumer subscription instead of a metered API key, which is usually the cheapest way to run gptme on a frontier model:

  • ChatGPT Plus/Pro (Codex) — openai-subscription/gpt-6-astra (and the GPT-5.6 family). Authenticate once with gptme-auth openai-subscription.

  • SuperGrok (Grok Build) — grok-subscription/grok-4.6. Reuses the grok CLI’s login, or gptme-auth grok-subscription.

  • Claude Max — not available: Anthropic does not permit third-party tools on the consumer subscription; use the anthropic provider with an API key.

  • Cursor — not supported. Cursor exposes no sanctioned endpoint for using a personal subscription from third-party tools (its Cloud Agents API is separately metered and is not a chat-completions API). Unofficial proxies that reuse the Cursor CLI’s login state exist, but gptme does not integrate them.

See OpenAI Subscription and Grok Subscription for setup details.

Configuring credentials#

To configure provider credentials interactively, run /account inside gptme:

/account
/account setup
/account setup openrouter

/account setup openrouter starts browser-based OpenRouter sign-in using OAuth / PKCE, stores the resulting key in ~/.config/gptme/credentials.toml, and switches the default model to OpenRouter’s recommended default.

For providers without OAuth onboarding yet, /account setup <provider> prompts for the key without putting it in shell history and stores it in ~/.config/gptme/credentials.toml (or $XDG_CONFIG_HOME/gptme/credentials.toml if set). Supported manual providers currently include anthropic, openai, deepseek, gemini, groq, and xai.

Subscription sign-in (ChatGPT Plus/Pro and SuperGrok) is also offered in the first-run setup when gptme starts without any configured credentials.

You can still use the [env] section in the Global config file to store API keys using the same format as the environment variables:

  • OPENAI_API_KEY="your-api-key"

  • ANTHROPIC_API_KEY="your-api-key"

  • OPENROUTER_API_KEY="your-api-key"

  • GEMINI_API_KEY="your-api-key"

  • XAI_API_KEY="your-api-key"

  • GROQ_API_KEY="your-api-key"

  • DEEPSEEK_API_KEY="your-api-key"

Reasoning effort#

Reasoning models accept a named effort level that trades latency and cost for more thinking. gptme exposes one knob, GPTME_THINKING_EFFORT, and maps it to each provider’s parameter:

Provider

Request parameter

Accepted levels

Anthropic

thinking.budget_tokens (and output_config.effort on SDK >= 0.77)

low, medium, high, xhigh, max

OpenAI (Chat Completions)

reasoning_effort

none, minimal, low, medium, high, xhigh, max

OpenAI (Responses API)

reasoning.effort

same as above

OpenRouter

reasoning.effort (replaces the default reasoning.max_tokens budget)

none, minimal, low, medium, high, xhigh

Moonshot Kimi K3

reasoning_effort

low, high, max

OpenAI Subscription (Codex)

reasoning.effort from the model :level suffix (default medium)

low, medium, high, xhigh

Which subset a specific model accepts (for example none on gpt-5.1+, xhigh on gpt-5.2+) is enforced by the provider; gptme only rejects levels the provider never accepts. Models without reasoning support ignore the variable, so it is safe to leave set across model switches.

GPTME_THINKING_EFFORT=high gptme "prove this" -m openai/gpt-5.5
GPTME_THINKING_EFFORT=low gptme "rename the variable" -m openrouter/deepseek/deepseek-r1

Every assistant message records what happened, so session logs are not blind to effort:

  • metadata.reasoning_effort - the level that shaped the request (only set when one applied; absent means the provider default).

  • metadata.usage.reasoning_tokens - reasoning tokens reported by the provider (OpenAI, OpenRouter, Codex). Anthropic bills thinking inside output_tokens and does not report it separately.

Custom, local, and plugin providers#

Any OpenAI-compatible server (Ollama, LM Studio, vLLM, a private proxy) can be used with the local/ prefix or declared as a named [[providers]] entry in your config — see Custom and Local Providers.

Third-party packages can also register providers through the gptme.providers entry point, making them available right after installation:

pip install gptme-provider-minimax
gptme "hello" -m minimax/MiniMax-M3

See Provider Integration Guide to write one.

Contents