Models
Model setups
Results f24d1f41, updated 16 November 2026, 22:20 UTC; scored by GNU Backgammon 1.08, 2-ply.
League 1
FinishedPublishedSwipe for more columns →
| Setup | Model type | Weights | Runs on | Reasoning setting | Confidence source | Missing B7 fields |
|---|---|---|---|---|---|---|
| Perplexity Decider 27B v1.1NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkdecision model · our GPUs | decision model | open | our GPUs | none (decision model) | decision-model probability | 1 |
| SubQ Snap Previewdecision model · API | decision model | Missing | API | none (decision model) | decision-model probability | 2 |
| Open-Jev-27B-v1.1NVIDIA RTX PRO 6000 Blackwelldecision model · our GPUs | decision model | open | our GPUs | none (decision model) | decision-model probability | 0 |
| CygnetNVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkdecision model · our GPUs | decision model | open | our GPUs | none (letter read directly) | letter probability | 1 |
| Randomprogram · CPU | program | Missing | CPU | Missing | none | 0 |
| Safestprogram · CPU | program | Missing | CPU | Missing | none | 0 |
| GNUbg weakprogram · CPU | program | Missing | CPU | Missing | none | 0 |
| Jev 1.13decision model · API | decision model | Missing | API | none (decision model) | decision-model probability | 1 |
| GPT-6 LunaDecisions APIdecision model · API | decision model | Missing | API | none (decision model) | decision-model probability | 1 |
| Perplexity Decider Flash v1.2NVIDIA RTX PRO 6000 Blackwelldecision model · our GPUs | decision model | open | our GPUs | none (decision model) | decision-model probability | 0 |
| Winnow-12BNVIDIA RTX PRO 6000 Blackwelldecision model · our GPUs | decision model | open | our GPUs | none (decision model) | decision-model probability | 0 |
| Mercury Decidedecision model · API | decision model | Missing | API | none (decision model) | decision-model probability | 1 |
| Perplexity Decider 27B v1.2NVIDIA RTX PRO 6000 Blackwelldecision model · our GPUs | decision model | open | our GPUs | none (decision model) | decision-model probability | 0 |
| GLiDEv1.2 · NVIDIA RTX PRO 6000 Blackwelldecision model · our GPUs | decision model | open | our GPUs | none (decision model) | decision-model probability | 1 |
| LFM2.5-24B-A2Boff · NVIDIA RTX PRO 6000 Blackwell | Missing | Missing | Missing | Missing | Missing | 1 |
| LFM2.5-24B-A2Bon · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Spark | Missing | Missing | Missing | Missing | Missing | 1 |
| Qwen3.8-27Boff · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Spark | Missing | Missing | Missing | Missing | Missing | 1 |
| DeepSeek-V4.1-Flashoff · API + NVIDIA RTX PRO 6000 Blackwell · first release | Missing | Missing | Missing | Missing | Missing | 1 |
| celeris-1-decisiondecision model · API | decision model | Missing | API | none (decision model) | decision-model probability | 1 |
League 2
FinishedPublishedSwipe for more columns →
| Setup | Model type | Weights | Runs on | Reasoning setting | Confidence source | Missing B7 fields |
|---|---|---|---|---|---|---|
| Gemini 3.8 Flashhighchat model, one letter · subscription | chat model, one letter | closed | subscription | thinking level high | none | 1 |
| Claude Sonnet 5.5lowchat model, one letter · subscription | chat model, one letter | closed | subscription | effort low | none | 1 |
| Claude Sonnet 5.5highchat model, one letter · subscription | chat model, one letter | closed | subscription | effort high | none | 1 |
| DeepSeek-V4.1-Flashlow · API + NVIDIA RTX PRO 6000 Blackwell · first releasechat model, answer-letter probabilities · API | chat model, answer-letter probabilities | open | API | thinking on, reasoning effort low | letter probability | 1 |
| GLM-5.3lowchat model, answer-letter probabilities · API | chat model, answer-letter probabilities | open | API | reasoning effort low (reasoning is mandatory on this endpoint, so no off setup) | letter probability | 1 |
| GNUbg weakprogram · CPU | program | Missing | CPU | Missing | none | 0 |
| Qwen3.8 Flashoffchat model, answer-letter probabilities · API | chat model, answer-letter probabilities | Missing | API | reasoning off | letter probability | 2 |
| Gemma 4 31Bon · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkchat model, answer-letter probabilities · our GPUs | chat model, answer-letter probabilities | open | our GPUs | thinking on (template default; no effort setting) | letter probability | 0 |
| Gemma 4 31Boff · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkchat model, answer-letter probabilities · our GPUs | chat model, answer-letter probabilities | open | our GPUs | thinking off | letter probability | 0 |
| Qwen3.8-27Blow · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkchat model, answer-letter probabilities · our GPUs | chat model, answer-letter probabilities | open | our GPUs | thinking on, reasoning effort low (the chat template's reasoning_effort) | letter probability | 0 |
| d1-ultra | Missing | Missing | Missing | Missing | Missing | 1 |
| Claude Fable 5.1low | Missing | Missing | Missing | Missing | Missing | 1 |
| GLiDEAPI | Missing | Missing | Missing | Missing | Missing | 1 |
| Grok 4.7low | Missing | Missing | Missing | Missing | Missing | 1 |
| DeepSeek-V4.1-Flashoff · API · 1017chat model, answer-letter probabilities · API | chat model, answer-letter probabilities | open | API | thinking off (reasoning disabled) | letter probability | 0 |
| DeepSeek-V4.1-Flashlow · API · 1017chat model, answer-letter probabilities · API | chat model, answer-letter probabilities | open | API | thinking on, reasoning effort low | letter probability | 0 |
| Jev 1.14 | Missing | Missing | Missing | Missing | Missing | 1 |
| ClefNVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Spark | Missing | Missing | Missing | Missing | Missing | 1 |
| Gemini 3.8 Flashlowchat model, one letter · subscription | chat model, one letter | closed | subscription | thinking level low | none | 1 |
League 3
FinishedPublishedSwipe for more columns →
| Setup | Model type | Weights | Runs on | Reasoning setting | Confidence source | Missing B7 fields |
|---|---|---|---|---|---|---|
| GPT-6 Astralowchat model, one letter · subscription | chat model, one letter | closed | subscription | reasoning effort low, the lowest gpt-6-astra accepts (none and minimal are rejected) | none | 1 |
| GPT-6 Astramediumchat model, one letter · subscription | chat model, one letter | closed | subscription | reasoning effort medium, the model's default | none | 1 |
| Claude Opus 5.5lowchat model, one letter · subscription | chat model, one letter | closed | subscription | effort low, the lowest Opus 5.5 accepts (adaptive thinking only) | none | 1 |
| Claude Opus 5.5highchat model, one letter · subscription | chat model, one letter | closed | subscription | effort high, the default (adaptive thinking) | none | 1 |
| Claude Fable 5.1highchat model, one letter · subscription | chat model, one letter | closed | subscription | effort high, the default (adaptive thinking) | none | 1 |
| Gemini 4 Argonhighchat model, one letter · subscription | chat model, one letter | closed | subscription | thinking level high, Gemini's default | none | 1 |
| Grok 4.7highchat model, one letter · subscription | chat model, one letter | closed | subscription | reasoning effort high, grok-4.7's default | none | 1 |
| GNUbg weakprogram · CPU | program | Missing | CPU | Missing | none | 0 |
| GPT-6.1 Sollow | Missing | Missing | Missing | Missing | Missing | 1 |
| GPT-6.1 Solhigh | Missing | Missing | Missing | Missing | Missing | 1 |
| Kimi K3.1 Prooffchat model, one letter · API | chat model, one letter | Missing | API | reasoning off: every request sent reasoning_effort "none" (the lowest effort OpenRouter lists for this model on 2026-10-24); every response reported 0 reasoning tokens | none | 1 |
| Gemini 4 Argonlowchat model, one letter · subscription | chat model, one letter | closed | subscription | thinking level low, the lowest | none | 1 |
v1 set
Earlier versionSwipe for more columns →
| Setup | Model type | Weights | Runs on | Reasoning setting | Confidence source | Missing B7 fields |
|---|---|---|---|---|---|---|
| Claude Opus 5.5chat model, one letter · subscription | chat model, one letter | closed | subscription | effort low | none | 3 |
| Clefdecision model · our GPUs | decision model | open | our GPUs | none (decision model) | decision-model probability | 1 |
| Cygnetdecision model · our GPUs | decision model | open | our GPUs | none (letter read directly) | letter probability | 2 |
| Perplexity Decider 27B v1.1decision model · our GPUs | decision model | open | our GPUs | none (decision model) | decision-model probability | 2 |
| DeepSeek-V4.1-Flashfirst releasechat model, answer-letter probabilities · our GPUs | chat model, answer-letter probabilities | open | our GPUs | Missing | letter probability | 3 |
| DeepSeek-V4.1-Flashthinking · first releasechat model, one letter · our GPUs | chat model, one letter | open | our GPUs | thinking on, reasoning effort low | none | 2 |
| GLM-4.7-Flashchat model, answer-letter probabilities · our GPUs | chat model, answer-letter probabilities | open | our GPUs | Missing | letter probability | 3 |
| GLM-5.3-Flash NVFP4chat model, answer-letter probabilities | chat model, answer-letter probabilities | Missing | Missing | Missing | letter probability | 9 |
| GPT-6 Astrachat model, one letter · subscription | chat model, one letter | closed | subscription | reasoning effort low | none | 3 |
| GPT-6 LunaDecisions APIdecision model · API | decision model | Missing | API | none (decision model) | decision-model probability | 3 |
| Gemini 3.8 Flashchat model, one letter · API | chat model, one letter | closed | API | thinking level low | none | 2 |
| Gemini 4 Argonchat model, one letter · API | chat model, one letter | closed | API | thinking level low | none | 3 |
| Gemini 4 Argonshuffledchat model, one letter · API · shuffled play order | chat model, one letter | closed | API | thinking level low | none | 3 |
| Gemma 4 31Bchat model, answer-letter probabilities · our GPUs | chat model, answer-letter probabilities | open | our GPUs | none (letter read directly) | letter probability | 2 |
| Jev 1.13decision model · API | decision model | Missing | API | none (decision model) | decision-model probability | 3 |
| Open-Jev 27Bdecision model · our GPUs | decision model | open | our GPUs | none (decision model) | decision-model probability | 2 |
| Qwen3.8 Flash Nextchat model, answer-letter probabilities · our GPUs | chat model, answer-letter probabilities | open | our GPUs | Missing | letter probability | 3 |
New models: first look (smoke tests)
Swipe for more columns →
| Setup | Model type | Weights | Runs on | Reasoning setting | Confidence source | Missing B7 fields |
|---|---|---|---|---|---|---|
| Beam 1.0highchat model, answer-letter probabilities · API | chat model, answer-letter probabilities | open | API | reasoning effort high (reasoning is mandatory) | letter probability | 0 |
| Beam 1.0lowchat model, answer-letter probabilities · API | chat model, answer-letter probabilities | open | API | reasoning effort low (reasoning is mandatory) | letter probability | 0 |
| Claude Fable 5.1highchat model, one letter · subscription | chat model, one letter | closed | subscription | effort high (adaptive thinking only) | none | 0 |
| Claude Fable 5.1lowchat model, one letter · subscription | chat model, one letter | closed | subscription | effort low (adaptive thinking only) | none | 0 |
| Claude Haiku 5.5highchat model, one letter · subscription | chat model, one letter | closed | subscription | effort high | none | 0 |
| Claude Haiku 5.5lowchat model, one letter · subscription | chat model, one letter | closed | subscription | effort low | none | 0 |
| Gemini 3.8 Flashlow · antigravitychat model, one letter · subscription | chat model, one letter | closed | subscription | thinking level low | none | 0 |
| Grok 4.7highchat model, one letter · subscription | chat model, one letter | closed | subscription | reasoning effort high, the default | none | 0 |
| Grok 4.7lowchat model, one letter · subscription | chat model, one letter | closed | subscription | reasoning effort low | none | 0 |
| Hy4 Previewoffchat model, one letter · API | chat model, one letter | open | API | reasoning off | none | 0 |
| Kimi K3lowchat model, one letter · API | chat model, one letter | open | API | effort low (an effort is required; no off) | none | 0 |
| Mercury 3lowchat model, one letter · API | chat model, one letter | closed | API | reasoning effort low | none | 0 |
| Mercury 3offchat model, one letter · API | chat model, one letter | closed | API | reasoning off | none | 0 |
| MiMo-V2.6-Flashlowchat model, answer-letter probabilities · API | chat model, answer-letter probabilities | open | API | reasoning effort low | letter probability | 0 |
| MiMo-V2.6-Flashoffchat model, answer-letter probabilities · API | chat model, answer-letter probabilities | open | API | reasoning off | letter probability | 0 |
| MiMo-V2.6-Flashoff, unconstrainedchat model, answer-letter probabilities · API | chat model, answer-letter probabilities | open | API | reasoning off | letter probability | 0 |
| MiMo V2.6 Prohighchat model, one letter · API | chat model, one letter | open | API | reasoning effort high | none | 0 |
| MiMo V2.6 Prooffchat model, one letter · API | chat model, one letter | open | API | reasoning off | none | 0 |
| MiniMax-M3.1-Flashlowchat model, one letter · API | chat model, one letter | closed | API | effort low (reasoning is mandatory) | none | 0 |
| Nemotron 3.5 Boltoff | Missing | Missing | Missing | Missing | Missing | 1 |
| Nemotron 3.5 Bolton | Missing | Missing | Missing | Missing | Missing | 1 |
| d3-turbo | Missing | Missing | Missing | Missing | Missing | 1 |
About these tables
One page per setup, with how it was called. Setups are not identical: they differ in model type, where they run, reasoning setting, confidence source and option groups.