LLM Providers (BYOLLM)
Neural Inverse never routes your code through Neural Inverse servers. You bring your own API keys — they are stored locally and used directly from your machine.
Supported Providers
Cloud Providers
| Provider | Setup |
|---|---|
| Anthropic | API key from console.anthropic.com |
| OpenAI | API key from platform.openai.com |
| Google Gemini | API key from aistudio.google.com |
| xAI (Grok) | API key from x.ai |
| DeepSeek | API key from platform.deepseek.com |
| Mistral | API key from console.mistral.ai |
| Groq | API key from console.groq.com |
| OpenRouter | API key from openrouter.ai |
| GitHub Models | GitHub PAT with models:read scope |
| Fireworks AI | API key from fireworks.ai |
| Cerebras | API key from cloud.cerebras.ai |
| AWS Bedrock | AWS credentials + region (default: us-east-1) |
| Google Vertex AI | GCP project + region (default: us-west2) |
| Microsoft Azure | Azure resource name + API key + API version |
Local / Self-Hosted Providers
| Provider | Default endpoint | Notes |
|---|---|---|
| Ollama | http://localhost:11434 | Models auto-detected |
| vLLM | http://localhost:8000 | Models auto-detected |
| LM Studio | http://localhost:1234 | Models auto-detected |
| LiteLLM | Custom endpoint | OpenAI-compatible proxy |
| OpenAI-Compatible | Custom endpoint | Any OpenAI-format API, custom headers supported |
Default Models
Neural Inverse ships with a curated default model list per provider. You can add any model string not in the list.
Anthropic: claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5-20251001, claude-3-7-sonnet-latest, claude-3-5-sonnet-latest, claude-3-5-haiku-latest
OpenAI: gpt-5.4, gpt-5.4-mini, gpt-5.4-nano, gpt-5.1-codex, o3, o4-mini
Google Gemini: gemini-3.1-pro-preview, gemini-3-flash-preview, gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite, gemini-2.0-flash
xAI: grok-4.20, grok-4-1-fast-reasoning, grok-3, grok-3-mini
DeepSeek: deepseek-chat, deepseek-reasoner
GitHub Models: openai/gpt-4.1, openai/gpt-4.1-mini, openai/gpt-4.1-nano, openai/o4-mini, openai/o3-mini, deepseek/deepseek-r1, meta/llama-4-scout-17b-16e-instruct, mistralai/mistral-small-2503, xai/grok-3-mini
Fireworks AI: accounts/fireworks/models/llama-v3p3-70b-instruct, accounts/fireworks/models/deepseek-r1, accounts/fireworks/models/qwen3-235b-a22b, accounts/fireworks/models/qwen3-32b, accounts/fireworks/models/gemma-4-31b-it, accounts/fireworks/models/gpt-oss-120b, accounts/fireworks/models/gpt-oss-20b
Cerebras: llama3.1-8b, gpt-oss-120b, qwen-3-235b-a22b-instruct-2507
Ollama / vLLM / LM Studio: Models are auto-detected from your running server — no manual list needed.
Adding a Provider
- Open Settings (gear icon or
Cmd+,/Ctrl+,). - Go to Neural Inverse > LLM Providers.
- Select the provider and enter your API key or endpoint.
- Click Verify to test the connection.
For local providers (Ollama, vLLM, LM Studio), start your server first — Neural Inverse polls the endpoint to discover available models automatically.
Selecting a Model per Feature
Neural Inverse lets you assign different models to different features:
- Chat — the model used in the sidebar chat panel
- Autocomplete — the model used for inline code completions
- Ctrl+K — the model used for inline edit (Cmd+K / Ctrl+K)
- Apply — the model used when applying suggested changes
- Power Mode — the model used by Power Mode agents
Each feature has an independent model selection. You can use a fast local model for autocomplete and a powerful cloud model for Power Mode.
OpenAI-Compatible Endpoints
For any provider that exposes an OpenAI-format API (LiteLLM proxy, custom inference server, etc.):
- Select OpenAI Compatible as the provider.
- Enter the base endpoint URL.
- Enter the API key (leave blank if not required).
- Optionally add custom request headers as JSON (e.g. for authentication headers).
Air-Gapped / Offline Setup
For environments with no internet access:
- Run Ollama or vLLM on a local server within your network.
- Point Neural Inverse at the local endpoint.
- All LLM traffic stays within your network boundary.
No Neural Inverse cloud dependency is required after initial installation.