Models

Model setups

Results f24d1f41, updated 16 November 2026, 22:20 UTC; scored by GNU Backgammon 1.08, 2-ply.

League 1

FinishedPublished

Swipe for more columns →

Setups in this league
SetupModel typeWeightsRuns onReasoning settingConfidence sourceMissing B7 fields
Perplexity Decider 27B v1.1NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkdecision model · our GPUsdecision modelopenour GPUsnone (decision model)decision-model probability1
SubQ Snap Previewdecision model · APIdecision modelMissingAPInone (decision model)decision-model probability2
Open-Jev-27B-v1.1NVIDIA RTX PRO 6000 Blackwelldecision model · our GPUsdecision modelopenour GPUsnone (decision model)decision-model probability0
CygnetNVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkdecision model · our GPUsdecision modelopenour GPUsnone (letter read directly)letter probability1
Randomprogram · CPUprogramMissingCPUMissingnone0
Safestprogram · CPUprogramMissingCPUMissingnone0
GNUbg weakprogram · CPUprogramMissingCPUMissingnone0
Jev 1.13decision model · APIdecision modelMissingAPInone (decision model)decision-model probability1
GPT-6 LunaDecisions APIdecision model · APIdecision modelMissingAPInone (decision model)decision-model probability1
Perplexity Decider Flash v1.2NVIDIA RTX PRO 6000 Blackwelldecision model · our GPUsdecision modelopenour GPUsnone (decision model)decision-model probability0
Winnow-12BNVIDIA RTX PRO 6000 Blackwelldecision model · our GPUsdecision modelopenour GPUsnone (decision model)decision-model probability0
Mercury Decidedecision model · APIdecision modelMissingAPInone (decision model)decision-model probability1
Perplexity Decider 27B v1.2NVIDIA RTX PRO 6000 Blackwelldecision model · our GPUsdecision modelopenour GPUsnone (decision model)decision-model probability0
GLiDEv1.2 · NVIDIA RTX PRO 6000 Blackwelldecision model · our GPUsdecision modelopenour GPUsnone (decision model)decision-model probability1
LFM2.5-24B-A2Boff · NVIDIA RTX PRO 6000 BlackwellMissingMissingMissingMissingMissing1
LFM2.5-24B-A2Bon · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX SparkMissingMissingMissingMissingMissing1
Qwen3.8-27Boff · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX SparkMissingMissingMissingMissingMissing1
DeepSeek-V4.1-Flashoff · API + NVIDIA RTX PRO 6000 Blackwell · first releaseMissingMissingMissingMissingMissing1
celeris-1-decisiondecision model · APIdecision modelMissingAPInone (decision model)decision-model probability1

League 2

FinishedPublished

Swipe for more columns →

Setups in this league
SetupModel typeWeightsRuns onReasoning settingConfidence sourceMissing B7 fields
Gemini 3.8 Flashhighchat model, one letter · subscriptionchat model, one letterclosedsubscriptionthinking level highnone1
Claude Sonnet 5.5lowchat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort lownone1
Claude Sonnet 5.5highchat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort highnone1
DeepSeek-V4.1-Flashlow · API + NVIDIA RTX PRO 6000 Blackwell · first releasechat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesopenAPIthinking on, reasoning effort lowletter probability1
GLM-5.3lowchat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesopenAPIreasoning effort low (reasoning is mandatory on this endpoint, so no off setup)letter probability1
GNUbg weakprogram · CPUprogramMissingCPUMissingnone0
Qwen3.8 Flashoffchat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesMissingAPIreasoning offletter probability2
Gemma 4 31Bon · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkchat model, answer-letter probabilities · our GPUschat model, answer-letter probabilitiesopenour GPUsthinking on (template default; no effort setting)letter probability0
Gemma 4 31Boff · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkchat model, answer-letter probabilities · our GPUschat model, answer-letter probabilitiesopenour GPUsthinking offletter probability0
Qwen3.8-27Blow · NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Sparkchat model, answer-letter probabilities · our GPUschat model, answer-letter probabilitiesopenour GPUsthinking on, reasoning effort low (the chat template's reasoning_effort)letter probability0
d1-ultraMissingMissingMissingMissingMissing1
Claude Fable 5.1lowMissingMissingMissingMissingMissing1
GLiDEAPIMissingMissingMissingMissingMissing1
Grok 4.7lowMissingMissingMissingMissingMissing1
DeepSeek-V4.1-Flashoff · API · 1017chat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesopenAPIthinking off (reasoning disabled)letter probability0
DeepSeek-V4.1-Flashlow · API · 1017chat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesopenAPIthinking on, reasoning effort lowletter probability0
Jev 1.14MissingMissingMissingMissingMissing1
ClefNVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX SparkMissingMissingMissingMissingMissing1
Gemini 3.8 Flashlowchat model, one letter · subscriptionchat model, one letterclosedsubscriptionthinking level lownone1

League 3

FinishedPublished

Swipe for more columns →

Setups in this league
SetupModel typeWeightsRuns onReasoning settingConfidence sourceMissing B7 fields
GPT-6 Astralowchat model, one letter · subscriptionchat model, one letterclosedsubscriptionreasoning effort low, the lowest gpt-6-astra accepts (none and minimal are rejected)none1
GPT-6 Astramediumchat model, one letter · subscriptionchat model, one letterclosedsubscriptionreasoning effort medium, the model's defaultnone1
Claude Opus 5.5lowchat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort low, the lowest Opus 5.5 accepts (adaptive thinking only)none1
Claude Opus 5.5highchat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort high, the default (adaptive thinking)none1
Claude Fable 5.1highchat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort high, the default (adaptive thinking)none1
Gemini 4 Argonhighchat model, one letter · subscriptionchat model, one letterclosedsubscriptionthinking level high, Gemini's defaultnone1
Grok 4.7highchat model, one letter · subscriptionchat model, one letterclosedsubscriptionreasoning effort high, grok-4.7's defaultnone1
GNUbg weakprogram · CPUprogramMissingCPUMissingnone0
GPT-6.1 SollowMissingMissingMissingMissingMissing1
GPT-6.1 SolhighMissingMissingMissingMissingMissing1
Kimi K3.1 Prooffchat model, one letter · APIchat model, one letterMissingAPIreasoning off: every request sent reasoning_effort "none" (the lowest effort OpenRouter lists for this model on 2026-10-24); every response reported 0 reasoning tokensnone1
Gemini 4 Argonlowchat model, one letter · subscriptionchat model, one letterclosedsubscriptionthinking level low, the lowestnone1

v1 set

Earlier version

Swipe for more columns →

Setups of the earlier version
SetupModel typeWeightsRuns onReasoning settingConfidence sourceMissing B7 fields
Claude Opus 5.5chat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort lownone3
Clefdecision model · our GPUsdecision modelopenour GPUsnone (decision model)decision-model probability1
Cygnetdecision model · our GPUsdecision modelopenour GPUsnone (letter read directly)letter probability2
Perplexity Decider 27B v1.1decision model · our GPUsdecision modelopenour GPUsnone (decision model)decision-model probability2
DeepSeek-V4.1-Flashfirst releasechat model, answer-letter probabilities · our GPUschat model, answer-letter probabilitiesopenour GPUsMissingletter probability3
DeepSeek-V4.1-Flashthinking · first releasechat model, one letter · our GPUschat model, one letteropenour GPUsthinking on, reasoning effort lownone2
GLM-4.7-Flashchat model, answer-letter probabilities · our GPUschat model, answer-letter probabilitiesopenour GPUsMissingletter probability3
GLM-5.3-Flash NVFP4chat model, answer-letter probabilitieschat model, answer-letter probabilitiesMissingMissingMissingletter probability9
GPT-6 Astrachat model, one letter · subscriptionchat model, one letterclosedsubscriptionreasoning effort lownone3
GPT-6 LunaDecisions APIdecision model · APIdecision modelMissingAPInone (decision model)decision-model probability3
Gemini 3.8 Flashchat model, one letter · APIchat model, one letterclosedAPIthinking level lownone2
Gemini 4 Argonchat model, one letter · APIchat model, one letterclosedAPIthinking level lownone3
Gemini 4 Argonshuffledchat model, one letter · API · shuffled play orderchat model, one letterclosedAPIthinking level lownone3
Gemma 4 31Bchat model, answer-letter probabilities · our GPUschat model, answer-letter probabilitiesopenour GPUsnone (letter read directly)letter probability2
Jev 1.13decision model · APIdecision modelMissingAPInone (decision model)decision-model probability3
Open-Jev 27Bdecision model · our GPUsdecision modelopenour GPUsnone (decision model)decision-model probability2
Qwen3.8 Flash Nextchat model, answer-letter probabilities · our GPUschat model, answer-letter probabilitiesopenour GPUsMissingletter probability3

New models: first look (smoke tests)

Swipe for more columns →

Smoke-tested setups
SetupModel typeWeightsRuns onReasoning settingConfidence sourceMissing B7 fields
Beam 1.0highchat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesopenAPIreasoning effort high (reasoning is mandatory)letter probability0
Beam 1.0lowchat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesopenAPIreasoning effort low (reasoning is mandatory)letter probability0
Claude Fable 5.1highchat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort high (adaptive thinking only)none0
Claude Fable 5.1lowchat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort low (adaptive thinking only)none0
Claude Haiku 5.5highchat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort highnone0
Claude Haiku 5.5lowchat model, one letter · subscriptionchat model, one letterclosedsubscriptioneffort lownone0
Gemini 3.8 Flashlow · antigravitychat model, one letter · subscriptionchat model, one letterclosedsubscriptionthinking level lownone0
Grok 4.7highchat model, one letter · subscriptionchat model, one letterclosedsubscriptionreasoning effort high, the defaultnone0
Grok 4.7lowchat model, one letter · subscriptionchat model, one letterclosedsubscriptionreasoning effort lownone0
Hy4 Previewoffchat model, one letter · APIchat model, one letteropenAPIreasoning offnone0
Kimi K3lowchat model, one letter · APIchat model, one letteropenAPIeffort low (an effort is required; no off)none0
Mercury 3lowchat model, one letter · APIchat model, one letterclosedAPIreasoning effort lownone0
Mercury 3offchat model, one letter · APIchat model, one letterclosedAPIreasoning offnone0
MiMo-V2.6-Flashlowchat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesopenAPIreasoning effort lowletter probability0
MiMo-V2.6-Flashoffchat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesopenAPIreasoning offletter probability0
MiMo-V2.6-Flashoff, unconstrainedchat model, answer-letter probabilities · APIchat model, answer-letter probabilitiesopenAPIreasoning offletter probability0
MiMo V2.6 Prohighchat model, one letter · APIchat model, one letteropenAPIreasoning effort highnone0
MiMo V2.6 Prooffchat model, one letter · APIchat model, one letteropenAPIreasoning offnone0
MiniMax-M3.1-Flashlowchat model, one letter · APIchat model, one letterclosedAPIeffort low (reasoning is mandatory)none0
Nemotron 3.5 BoltoffMissingMissingMissingMissingMissing1
Nemotron 3.5 BoltonMissingMissingMissingMissingMissing1
d3-turboMissingMissingMissingMissingMissing1
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.