Allowance
The models Triplo provides with your licence, how many tokens you get each month, and the whole list as the model service returns it today.
Models that come with the licence
What they are. A set of models Triplo provides with your licence. No API key, no separate account, no per-token billing from you. They appear in the model selector under Allowance Models.
How to reach them. They are on by default. To turn them off — for instance if you only ever want your own provider's models in the list — open the action grid → LLM Models and use the toggle on the Allowance Models row.
What the allowance is for
It exists so that Triplo works the moment you install it: press the hotkey, ask something, get an answer, without an account somewhere else, an API key, or a card on file. That is the whole job.
It is deliberately not a way to run frontier models for free, and not a way around what tokens cost. The set is built from small and fast models, the monthly allocation is finite, and heavy or high-stakes work belongs on a Connection of your own, where you choose the model and pay your provider directly. Agents are the clearest case: Pi never runs on the allowance at all.
What is in the set
Triplo fetches the list when the app starts, so it changes as models are added and retired. These 18 vendors are in it at the time of writing — 37 models in total:
| Vendor | Models |
|---|---|
| Amazon | Nova Micro 1.0 |
| ByteDance Seed | Seed 1.6 Flash |
| DeepSeek | DeepSeek V4 Flash (the default pick), DeepSeek V3.2, DeepSeek V3.2 Exp |
| Gemini 2.5 Flash Lite, Gemma 4 31B, Gemma 4 26B A4B | |
| IBM | Granite 4.1 8B |
| InclusionAI | Ling-3.0-flash |
| Meta | Llama 4 Scout, Llama 4 Maverick, Llama 3.3 70B Instruct |
| Microsoft | Phi 4 |
| Mistral | Ministral 3 14B, Ministral 3 8B, Small 3.2 24B, Saba |
| Nex AGI | Nex-N2-Mini |
| NVIDIA | Nemotron 3.5 Lightning, Nemotron 3 Nano 30B A3B |
| OpenAI | GPT-5 Nano, GPT 4.1 Nano, GPT 4o-mini 128k, GPT OSS 120B, GPT OSS 20B |
| Poolside | Laguna S 2.1, Laguna XS 2.1 |
| Qwen | Qwen3.7 Flash, Qwen3 30B A3B Thinking 2507, Qwen3 32B, Qwen3 14B |
| Reka | Reka Edge |
| Tencent | Hy3 |
| Upstage | Solar Pro 4 |
| Z.ai | GLM 5.3 Flash, GLM 4.7 Flash |
Treat that as a snapshot, not a contract — the selector is the source of truth, and it is the only place that shows what your licence can reach today.
How much you get
The allowance is measured in tokens, counted per device, per month, and it resets when the month turns over. Two devices on the same licence each get their own allocation.
| Per device, per month | |
|---|---|
| Standard allocation | 2 million tokens |
| In effect now | 4 million tokens, through January 2027 |
The standard allocation is 2 million tokens. It is currently doubled to 4 million, and that stays in place until January 2027, when we will review it. Plan around the 2 million figure if you are deciding whether the allowance covers your work — the extra is a bonus, not a commitment.
The prompt bar shows what you have used against the total, so you never have to guess. A short question and answer costs a few hundred tokens; a long document pasted in with a detailed reply costs a few thousand. Grounding a prompt on a knowledge base adds the retrieved passages to that count.
Limits & gotchas
- Allowance models require an internet connection and a valid licence.
- When the monthly allocation runs out, allowance models stop until the month turns over — switch to a Connection or a local model to keep working.
- If the list is empty and shows an error, Triplo could not reach the model service. Check your connection and reopen the drawer.
Everything else about picking and mixing models is in Models & Connections.