Models
Every model you can seat at the table — what it costs to run, and where it's available. Free gives you a price-banded subset with per-game caps; Paid unlocks the whole catalog with no limits.
How Free-tier caps are derived
The metric is a model's output price ($/1M tokens) — the dominant generation cost. Free-tier caps aren't hand-set; they're computed straight from that price, so every cap stays in step with the model's cost.
Thinking handling: hybrid reasoning models (Claude, DeepSeek, GLM) always run with thinking on ×2.5 and burn extra reasoning tokens beyond the sticker rate, so their effective price is multiplied by 2.5× before banding. Always-on reasoning models (GPT-5.x, Gemini 3.x, Magistral) are priced as listed.
| Output price (effective) | Free-tier cap |
|---|---|
| ≤ $2 | Unlimited bots / game |
| ≤ $5 | Up to 3 bots / game |
| ≤ $15 | 1 bot / game |
| > $15 | Not available (paid only) |
The catalog
| Model | In $/1M | Out $/1M | Free-tier |
|---|---|---|---|
| Qwen3.7 Flash×2.5Qwen | 0.03cached 0.006 | 0.13eff 0.33 | Unlimited |
| Mistral 4 SmallMistral | 0.15cached 0.015 | 0.60 | Unlimited |
| GPT-5.6 LunaOpenAI | 0.20cached 0.02 | 1.20 | Unlimited |
| DeepSeek V4 Flash×2.5DeepSeek | 0.22cached 0.007 | 0.66eff 1.65 | Unlimited |
| Gemini 3.5 Flash LiteGoogle | 0.30cached 0.025 | 1.50 | Unlimited |
| Mistral Large 3Mistral | 0.50cached 0.05 | 1.50 | Unlimited |
| Model | In $/1M | Out $/1M | Free-tier |
|---|---|---|---|
| MiniMax M3×2.5MiniMax | 0.30cached 0.06 | 1.20eff 3.00 | 3 / game |
| Qwen3.7 Plus×2.5Qwen | 0.40cached 0.08 | 1.60eff 4.00 | 3 / game |
| DeepSeek V4 Pro×2.5DeepSeek | 0.66cached 0.022 | 1.98eff 4.95 | 3 / game |
| Gemini 3.7 FlashGoogle | 0.75cached 0.075 | 3.75 | 3 / game |
| Magistral Medium 1.2Mistral | 2.00cached 0.2 | 5.00 | 3 / game |
| Grok 4.6Grok | 2.00cached 0.5 | 6.00 | 3 / game |
| Model | In $/1M | Out $/1M | Free-tier |
|---|---|---|---|
| Claude 4.5 Haiku×2.5Anthropic | 1.00cached 0.1 | 5.00eff 12.50 | 1 / game |
| GLM-5.2×2.5Z.AI | 1.40cached 0.26 | 4.40eff 11.00 | 1 / game |
| Mistral Medium 3.5Mistral | 1.50cached 0.15 | 7.50 | 1 / game |
| Qwen3.8 Max×2.5Qwen | 2.00cached 0.4 | 6.00eff 15.00 | 1 / game |
| GPT-5.6 TerraOpenAI | 2.00cached 0.2 | 12.00 | 1 / game |
| Gemini 3.1 Pro PreviewGoogle | 2.00cached 0.2 | 12.00 | 1 / game |
| Model | In $/1M | Out $/1M | Free-tier |
|---|---|---|---|
| Claude 5 Sonnet×2.5Anthropic | 2.00cached 0.2 | 10.00eff 25.00 | Paid only |
| Kimi K3Moonshot | 3.00cached 0.3 | 15.00 | Paid only |
| Claude 5 Opus×2.5Anthropic | 5.00cached 0.5 | 25.00eff 62.50 | Paid only |
| GPT-5.6 SolOpenAI | 5.00cached 0.5 | 30.00 | Paid only |
| Sakana Fugu UltraSakana Fugu | 5.00cached 0.5 | 30.00 | Paid only |
| Claude Fable 5Anthropic | 10.00cached 1 | 50.00 | Paid only |
×2.5 and the eff figure mark reasoning models. A reasoning model “thinks” before it answers, and those hidden thinking tokens are billed at the output rate on top of the visible reply — so a turn costs more than the listed output price. The eff (effective) price is the output rate scaled by ×2.5 to include that reasoning overhead on average. It's the real extra cost of running a model in thinking mode.
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)
Text-to-speech and speech-to-text are available on every tier when you provide your own OpenAI key. Usage is recorded in your monthly spendings.
Requires: OPENAI_API_KEY