Models
Every model you can seat at the table — what it costs to run, and where it's available. Free gives you a cost-banded subset with per-game caps; Paid unlocks the whole catalog with no limits.
How Free-tier caps are derived
The metric is what one bot turn costs: the context the model reads, the reply it writes and the hidden thinking it does first, all billed. Free-tier caps aren't hand-set; they're computed straight from that number, so every cap stays in step with the model's real cost.
Why not the sticker price: every model here reasons before it answers, and how much varies a lot. Two models with the same output rate can differ five-fold in tokens written per turn, so the per-turn figure is measured from real games over the last 30 days (as of 2026-09-20) and re-measured periodically. A model that has not been played enough yet is banded on an estimate from its sticker prices until it has.
| Cost per turn | Free-tier cap |
|---|---|
| ≤ 0.3¢ | Unlimited bots / game |
| ≤ 1¢ | Up to 3 bots / game |
| ≤ 2¢ | 1 bot / game |
| > 2¢ | Not available (paid only) |
The catalog
| Model | In $/1M | Out $/1M | Per turn | Free-tier |
|---|---|---|---|---|
| Qwen3.8 FlashQwen | 0.15cached 0.016 | 0.47 | 0.08¢372 turns | Unlimited |
| Mistral 4 SmallMistral | 0.15cached 0.015 | 0.60 | 0.09¢1,021 turns | Unlimited |
| GLM-5.3 FlashZ.AI | 0.15cached 0.03 | 0.50 | 0.1¢481 turns | Unlimited |
| DeepSeek V4.1 FlashDeepSeek | 0.15cached 0.003 | 0.60 | 0.12¢539 turns | Unlimited |
| GPT-6 LunaOpenAI | 0.10cached 0.01 | 0.50 | 0.13¢1,011 turns | Unlimited |
| Gemini 3.5 Flash LiteGoogle | 0.30cached 0.025 | 1.50 | 0.18¢1,162 turns | Unlimited |
| MiniMax M3MiniMax | 0.30cached 0.06 | 1.20 | 0.26¢689 turns | Unlimited |
| Model | In $/1M | Out $/1M | Per turn | Free-tier |
|---|---|---|---|---|
| DeepSeek V4 ProDeepSeek | 0.66cached 0.022 | 1.98 | 0.58¢633 turns | 3 / game |
| GLM-5.3Z.AI | 1.40cached 0.26 | 4.40 | 0.75¢382 turns | 3 / game |
| Gemini 3.8 FlashGoogle | 0.75cached 0.075 | 3.75 | 0.92¢701 turns | 3 / game |
| Model | In $/1M | Out $/1M | Per turn | Free-tier |
|---|---|---|---|---|
| Mistral Medium 3.5Mistral | 1.50cached 0.15 | 7.50 | 1.1¢571 turns | 1 / game |
| Claude 4.5 HaikuAnthropic | 1.00cached 0.1 | 5.00 | 1.14¢451 turns | 1 / game |
| Qwen3.8 MaxQwen | 2.00cached 0.25 | 6.00 | 1.2¢426 turns | 1 / game |
| Muse Spark 1.3Meta | 1.25cached 0.15 | 4.25 | 1.41¢692 turns | 1 / game |
| GPT-5.6 TerraOpenAI | 2.00cached 0.2 | 12.00 | 1.61¢709 turns | 1 / game |
| Model | In $/1M | Out $/1M | Per turn | Free-tier |
|---|---|---|---|---|
| Sakana Fugu MaxSakana Fugu | 2.00cached 0.25 | 6.00 | 2.2¢estimate | Paid only |
| Claude 5.5 SonnetAnthropic | 2.00cached 0.2 | 10.00 | 2.6¢estimate | Paid only |
| GPT-6 SolOpenAI | 2.00cached 0.2 | 10.00 | 2.6¢estimate | Paid only |
| Gemini 3.1 Pro PreviewGoogle | 2.00cached 0.2 | 12.00 | 2.79¢667 turns | Paid only |
| Grok 4.7Grok | 2.00cached 0.5 | 6.00 | 3.54¢688 turns | Paid only |
| Kimi K3Moonshot | 3.00cached 0.3 | 15.00 | 3.9¢estimate | Paid only |
| Claude 5.5 OpusAnthropic | 4.00cached 0.2 | 20.00 | 5.2¢estimate | Paid only |
| Sakana Fugu UltraSakana Fugu | 5.00cached 0.5 | 30.00 | 7¢estimate | Paid only |
| Claude Fable 5.1Anthropic | 10.00cached 0.25 | 50.00 | 13¢estimate | Paid only |
| GPT-6 AstraOpenAI | 10.00cached 1 | 50.00 | 13¢estimate | Paid only |
Per turn is what one bot reply costs on average: the conversation context it reads, the reply it writes, and the hidden thinking tokens a reasoning model burns before answering, all billed at the listed rates. Measured over the last 30 days of real games (as of 2026-09-20); a model that has not been played enough yet shows an estimate from its sticker prices. It is the number the Free-tier caps are derived from, because two models with the same output price can differ five-fold in how much they write per turn.
On the Paid tier the entire catalog above is available with no per-game bot caps — the Free-tier column shows the limit that applies when you play for free.
Voice models (TTS / STT)
Each game is cast from one voice set, picked on the new-game form: OpenAI or Gemini. The set decides both how characters speak and which model transcribes your microphone. Voice runs on platform keys on every tier; usage is recorded in your monthly spendings.
Image models
Every new game gets a themed visual set generated once at creation: portrait avatars for the whole cast (including the Game Master) plus day and night scene illustrations for the chat — all drawn in a single style matching your game's theme. Eventful nights also get a one-off illustration of what happened, with the involved characters drawn from their portraits.