Models & Providers
Kognar Chat connects to multiple AI providers, giving you access to dozens of language models — all from one interface.
In Settings
Open Settings → Models & Providers to see every provider included with your account. Each card shows how many models that provider offers and an Active badge when at least one is available. Click a provider to turn that provider off, or to turn off only some of its models. Anything you turn off leaves the chat model list. The card stays in Settings so you can turn it back on.
Add Provider connects a model running locally with Ollama. Enter the address (usually http://127.0.0.1:11434), choose the models Ollama already has, and they show up in the chat selector under Ollama. The conversation then runs on this computer. Refresh models on that card asks Ollama again for what is installed and adds any new models to the chat list.
Model routing has two switches — Smart model routing and Auto-fallback on failure. Save Changes stores them on this device. The model you pick in the message field is still the one used to reply. Default models for new chats stay in Settings → General → Default models.
Selecting a model
The model selector lives in the bar directly below the message text, next to the Send button — not in the chat header. Click it to browse available models. You can filter by category:
| Filter | Use case |
|---|---|
| All | Show every available model |
| General | Everyday tasks and general-purpose use |
| Code | Programming, debugging, and code review |
| Image | Image generation and editing |
| Research | Web-connected and analysis-focused models |
Each model entry shows:
- Name and provider
- Context window size (how many tokens it can process)
- Tool support — whether the model can use agents and tools; models whose tool use has not been verified are labelled accordingly
You can only change the model before sending the first message in a conversation. Start a new conversation to use a different model.
The selector popover opens above the rest of the conversation so it doesn't get clipped inside the input.
Kognar only lists models your account can actually use in chat. Models marked inactive on the service — including image models — are hidden from the selector, the image modals and the default-model pickers.
If you pick a model that has no chat capability, or one that cannot use tools while agents are enabled, Kognar explains the mismatch instead of failing on send.
Default model
Set your preferred default model in Settings → General → Default models. This model will be pre-selected every time you start a new conversation.
You can set separate defaults for:
- Chat model — used for text conversations
- Image generation model — used when generating images
- Image editing model — used when editing existing images
If a saved default becomes unavailable in your account, Kognar switches to another active model automatically.
Supported providers
Models are grouped by provider in the selector. The groups appear in a fixed priority order — OpenAI, Google, Anthropic, Kognar, xAI, DeepSeek, Moonshot, Meta, Z.ai, Bedrock, Qwen — with an alphabetical fallback for any new group, so the most-used providers stay easy to find as the catalog grows.
| Provider | Examples in the catalog |
|---|---|
| OpenAI | GPT-6 Astra, GPT-5.5 / 5.5 Pro, GPT-5.4, GPT-5.1 Codex Max, gpt-oss-120b, GPT Image 1.5 / 2 |
| Gemini 3.5 Flash, Gemini 3.5 Flash Lite, Gemini 3.1 Pro, Gemini 3.7 Flash, Gemini 2.5 Flash, plus Gemini image models | |
| Anthropic | Claude Opus 4.7 / 4.6, Claude Sonnet 4.6, Claude Haiku 4.5 |
| Kognar | Kognar Engine, Kognar Engine 2.0, Kognar Engine Mini, Kognar Engine Chat |
| xAI | Grok 4.3, Grok 4.20, Grok 4.1 Fast |
| DeepSeek | DeepSeek v3.1 |
| Moonshot | Kimi K2.6, Kimi K2.5, Kimi K2 Thinking |
| Meta | Muse Spark 1.3 |
| Z.ai | GLM 5.1, GLM 5 |
| Qwen | Qwen 3.6 Plus, Qwen3 Next 80B, Qwen3 VL 30B Thinking |
| Blackforest Labs | FLUX.2 Pro, FLUX 1.1 Pro |
The exact list depends on what your account has access to — the selector shows what your key can use.
Recently added
| Model | Provider | Why you'd pick it |
|---|---|---|
| GPT-6 Astra | OpenAI | Very large context (~1.1M tokens) for long documents and codebases |
| Muse Spark 1.3 | Meta | Reasoning model with a 1M-token context |
| Gemini 3.5 Flash Lite | The lighter, faster option in the Gemini 3.5 line |
All three support chat, code and search with tool use.
Model classification badges
Each model in the input area is tagged with a classification badge:
| Badge | Meaning |
|---|---|
| smart | High-capability model, best for complex tasks |
| fast | Optimized for speed and lower latency |
| thinking | Extended reasoning mode, slower but deeper |
Hover over the badge to see a description.
Context window management
Each model has a maximum context window. As your conversation grows, the context meter in the header fills up. When you approach the limit:
- Kognar estimates the token count before sending
- If the message would exceed the limit, a Context Overflow dialog appears
- You can choose to compress the conversation or cancel
Compression reduces the conversation size using AI summarization while preserving the key information. If the conversation is still too large after one compression, a second, more aggressive compression is applied automatically.
Image models
When you ask Kognar to generate or edit an image for the first time in a conversation, a model selection dialog appears. Your choice is remembered for that conversation.
Supported image capabilities:
- Generation — Create images from text descriptions
- Editing — Modify or transform existing images
Generated images can be saved via right-click: Save to Downloads or Save As....