Private Preview
This feature is in private preview: it's not ready for production use, and it may be briefly unavailable as we deploy updates. To get access, sign up here.
Neon AI Gateway serves models hosted by Databricks. Use short model IDs in the model field — for example, claude-sonnet-4-6, gpt-5-mini, gemini-2-5-flash. The databricks- prefixed form is also accepted. The Neon Console and most examples use the short form.
important
Models are hosted by Databricks and served through Neon AI Gateway. By using these models, you are responsible for complying with each provider's applicable terms of use. See Provider terms below.
note
Model availability may vary by region. The catalog expands over time, so check back for new additions.
Quick picks
Not sure which model to use? Start here.
| I want to... | Recommended model | Open-weight alternative |
|---|---|---|
| Build a general chat or reasoning app | claude-sonnet-4-6 | llama-4-maverick |
| Understand images | claude-sonnet-4-6 or gemini-2-5-flash | llama-4-maverick |
| Process audio or video | gemini-3-5-flash | — |
| Generate or review code | gpt-5-3-codex | gpt-oss-120b |
| Run extended or deep reasoning | claude-opus-4-7 or gpt-5-1 | qwen3-next-80b-a3b-instruct |
| Handle very long documents (1M+ tokens) | gemini-3-1-pro or claude-opus-4-7 | qwen35-122b-a10b |
| Keep costs low for lightweight tasks | claude-haiku-4-5 or gpt-5-4-nano | meta-llama-3-1-8b-instruct |
Rate limits
During the private preview, the following limits apply:
| Limit | Value |
|---|---|
| Per model (tokens per minute) | 200,000 TPM |
| All models combined (tokens per minute) | 1,100,000 TPM |
If you hit a limit, you'll receive a 429 Too Many Requests response. Requests resume when the rate limit window resets.
These limits apply to input tokens. Upstream output token limits (20,000 OTPM for most models) apply independently, so you can hit a 429 on output tokens without reaching the input limit. See Databricks Foundation Model API limits for details.
Pricing
Inference is free during the private preview. When billing begins, prices will match each provider's published list prices.
Available models
Endpoints are shown in short form. See Which endpoint to use for full paths and when to prefer each one.
Anthropic
| Model ID | Inputs | Context | Endpoints | Notes |
|---|---|---|---|---|
claude-opus-4-8 | text, image | — | chat/completions · anthropic/messages | Most capable Claude model |
claude-opus-4-7 | text, image | 1M | chat/completions · anthropic/messages | Extended thinking available |
claude-opus-4-6 | text, image | 1M | chat/completions · anthropic/messages | Adaptive thinking with max effort level |
claude-opus-4-5 | text, image | 200K | chat/completions · anthropic/messages | Extended thinking available |
claude-opus-4-1 | text, image | 200K | chat/completions · anthropic/messages | |
claude-sonnet-4-6 | text, image | — | chat/completions · anthropic/messages | Recommended. Instant + extended thinking |
claude-sonnet-4-5 | text, image | — | chat/completions · anthropic/messages | Instant + extended thinking |
claude-sonnet-4 | text, image | — | chat/completions · anthropic/messages | Instant + extended thinking |
claude-haiku-4-5 | text, image | — | chat/completions · anthropic/messages | Fast and cost-effective |
| Model ID | Inputs | Context | Endpoints | Notes |
|---|---|---|---|---|
gemini-3-5-flash | text, image, video, audio | — | chat/completions · gemini | |
gemini-3-1-pro | text, image, video, audio | 1M | chat/completions · gemini | |
gemini-3-1-flash-lite | text, image, video, audio | — | chat/completions · gemini | |
gemini-3-pro | text, image, video, audio | 1M | chat/completions · gemini | |
gemini-3-flash | text, image, video, audio | — | chat/completions · gemini | |
gemini-2-5-pro | text, image, video, audio | 1M | chat/completions · gemini | Deep Think Mode |
gemini-2-5-flash | text, image, video, audio | 1M | chat/completions · gemini | |
gemma-3-12b | text, image | 128K | chat/completions | Chat completions only |
Databricks
Open-weight models hosted by Databricks. Text input only. Chat completions only.
| Model ID | Context | Endpoints | Notes |
|---|---|---|---|
databricks-gpt-oss-120b | 128K | chat/completions | Recommended. Adjustable reasoning effort |
databricks-gpt-oss-20b | 128K | chat/completions |
Meta
| Model ID | Inputs | Context | Endpoints | Notes |
|---|---|---|---|---|
databricks-llama-4-maverick | text, image | — | chat/completions | Recommended. Llama 4 Community License · Acceptable Use |
databricks-meta-llama-3-3-70b-instruct | text | 128K | chat/completions | Llama 3.3 Community License · Acceptable Use |
databricks-meta-llama-3-1-8b-instruct | text | 128K | chat/completions | Llama 3.1 Community License · Acceptable Use |
Alibaba
Text input only. Chat completions only.
| Model ID | Context | Endpoints | Notes |
|---|---|---|---|
databricks-qwen3-next-80b-a3b-instruct | — | chat/completions | |
databricks-qwen35-122b-a10b | 256K | chat/completions | Always reasons before responding |
OpenAI
All OpenAI models have a 400K token context window and accept text and image inputs.
| Model ID | Endpoints | Notes |
|---|---|---|
gpt-5-4 | chat/completions · openai/responses | |
gpt-5-4-mini | chat/completions · openai/responses | |
gpt-5-4-nano | chat/completions · openai/responses | |
gpt-5-3-codex | openai/responses | Requires OpenAI Responses API |
gpt-5-2-codex | openai/responses | Requires OpenAI Responses API |
gpt-5-2 | chat/completions · openai/responses | |
gpt-5-1-codex-max | openai/responses | Requires OpenAI Responses API |
gpt-5-1-codex-mini | openai/responses | Requires OpenAI Responses API |
gpt-5-1 | chat/completions · openai/responses | Instant + Thinking modes |
gpt-5 | chat/completions · openai/responses | |
gpt-5-mini | chat/completions · openai/responses | |
gpt-5-nano | chat/completions · openai/responses |
Which endpoint to use
Most models work with the Chat completions endpoint. It is the recommended starting point and works with all providers. Use a provider-specific endpoint when required:
| Provider | Recommended endpoint | Notes |
|---|---|---|
| Anthropic | /ai-gateway/mlflow/v1/chat/completions | Use /ai-gateway/anthropic/v1/messages for extended thinking and prompt caching |
| OpenAI (most models) | /ai-gateway/mlflow/v1/chat/completions | Use /ai-gateway/openai/v1/responses for Responses API features |
| OpenAI (codex variants) | /ai-gateway/openai/v1/responses | These models require the Responses API and don't work with chat/completions |
| Google Gemini | /ai-gateway/mlflow/v1/chat/completions | Use /ai-gateway/gemini/v1beta/models/{model}:generateContent with the google-genai SDK |
| Google Gemma 3 12B | /ai-gateway/mlflow/v1/chat/completions | Chat completions only. Doesn't support the Gemini SDK endpoint |
| Meta, Databricks, Alibaba | /ai-gateway/mlflow/v1/chat/completions | Chat completions only |
Provider terms
Models are hosted by Databricks and served through Neon AI Gateway. You are responsible for complying with each provider's applicable terms of use.
| Provider | Terms |
|---|---|
| Anthropic | Anthropic Usage Policy |
| OpenAI | OpenAI Usage Policies |
| Google Gemini | Google Cloud Acceptable Use Policy · Google Generative AI Prohibited Use Policy |
| Google Gemma | Gemma Terms of Use · Gemma Prohibited Use Policy |
| Meta | Terms differ by Llama version. See the Notes column in the Meta models table. |
Need help?
Join our Discord Server to ask questions or see what others are doing with Neon. For paid plan support options, see Support.