The right model for every job, in one catalog.
Use models from leading providers in chat, agents, workflows, and the API from one place. Auto picks a suitable model for each request; bring your own keys if you want, and decide which models are open to whom.
No credit card required.
Auto
Example scenarioPick a request
Auto is reading the request
- FastClaude Haiku 3.5 · Gemini 2.5 Flash · Gemini 2.0 FlashRelative usage: 1/4
- BalancedGPT-5 mini · GPT-4.1 mini · GPT-4oRelative usage: 2/4
- FlagshipGPT-5 · GPT-4.1 · Claude Opus 4.1Relative usage: 3/4
- Reasoningo3 · o3-pro · o4-miniRelative usage: 4/4
Catalog
Stop hopping between provider portals — it is all in one list.
Models your workspace enables are picked from the same catalog in chat, agents, workflows, and the API. Filter by provider, model type, or name, and see the context window, list price, and whether you can use your own key on a single row.
- OpenAI
- Anthropic
- xAI
- DeepSeek
- Qwen
- GLM
Illustrative list prices for planning. Live rates and availability follow your workspace catalog and may change.
| Model | Provider | Tier | Context | Input / 1M | Output / 1M | BYOK |
|---|---|---|---|---|---|---|
Auto ajan-auto | ajan.io | Balanced | 256k | — | — | — |
GPT-5 gpt-5 | OpenAI | Flagship | 256k | $1.25 | $10.00 | Yes |
GPT-5 mini gpt-5-mini | OpenAI | Balanced | 256k | $0.25 | $2.00 | Yes |
GPT-4.1 gpt-4.1 | OpenAI | Flagship | 128k | $2.00 | $8.00 | Yes |
GPT-4.1 mini gpt-4.1-mini | OpenAI | Balanced | 128k | $0.40 | $1.60 | Yes |
GPT-4o gpt-4o | OpenAI | Balanced | 128k | $2.50 | $10.00 | Yes |
o3 o3 | OpenAI | Reasoning | 200k | $10.00 | $40.00 | Yes |
o3-pro o3-pro | OpenAI | Reasoning | 200k | $20.00 | $80.00 | Yes |
30 models · unit: per 1M tokens · illustrative prices
Know your model types
Pick the right point between speed and depth.
Every provider offers models in similar tiers. Fast models shine on short, frequent tasks, balanced models on everyday work, flagship models on complex text, and reasoning models on problems that need step-by-step thinking. Not sure? Start with Auto.
Fast
Rewrites, classification, short answers.
Balanced
Summaries, drafts, everyday analysis.
Flagship
Long documents, careful writing, code.
Reasoning
Multi-step analysis, comparisons, math.
- Embedding modelsFor semantic search in knowledge bases.
- Image generation modelsFor creating images in chat.
Version number
Usually signals capability; a newer version is more capable.
mini · flash · nano
A size suffix signals speed and lower usage.
Thinking
A mode where the model reasons step by step before answering.
Switch models mid-conversation
Gather ideas with a fast model, write the final text with a strong one.
In chat, the model picker is at hand on every message; the context is kept and only the model producing the answer changes. Agents and workflow steps get the model the job needs pinned to them, so everyone gets the same quality.
- ChatA picker per message and a personal default model.
- AgentsA model and creativity setting pinned to the agent.
- WorkflowsA separate model for every AI step.
Bring your own keys
Manage model cost under your own provider agreement.
By default, models are included in your plan and need no setup. With BYOK you connect your own provider keys: the provider bills you directly for the model, and on the ajan.io side only a fraction of the equivalent usage is drawn as AI Credits.
Included models
Teams that want to start fast
- No setup
- Catalog models ready to use
- Usage draws from AI Credits
Your own keys (BYOK)
Organizations that want vendor and cost control
- The provider bills you directly
- Only the models you configure
- Detailed cost and token breakdown
Same work, two ways
- With included models100 AI Credits
- With your own keys20 AI Credits
Illustrative. The provider’s charge goes to your own account separately.
With BYOK, configure at least one model of each type
- Completion modelFor chat and agents.
- Embedding modelFor document search and knowledge bases.
- Image generation modelFor creating images.
Fallback deployments for continuity
When one endpoint hits its limit, the model keeps answering.
A model you add with your own keys can have several deployments: a different region, endpoint, or key. Requests are spread across enabled deployments by their tokens-per-minute limits; if one reaches its limit or becomes unavailable, the request goes to another.
Which model, for whom, how much
Model access is a governance decision, not a personal preference.
Admins turn models on or off per workspace and keep expensive models for the roles that need them. Usage is visible per user, agent, and model, and usage limits and spend limits can be set. Users see their context window and remaining usage from the indicator in the conversation.
Model access
Enable models per workspace or group.
Usage visibility
Track by user, agent, and model.
Limits
Set usage and spend limits.
Context indicator
Users see their context and limit status.
Your data does not train models
Whichever model you pick, the rule for your content is the same.
No training
Customer prompts and files are not used to train foundation models.
Workspace boundary
Uploads and conversations are isolated per workspace.
Your own agreement
With BYOK, requests go through your provider account.
Questions about models
What people ask most about model choice and cost.
All FAQs
Auto reads your first message, estimates the request type and complexity, and picks a suitable model for that conversation. It avoids spending usage on powerful models for simple tasks.
Pick models per task, not per vendor portal.
Start free, then decide which models and keys your workspace should expose.