Aircade retro glove controller header

Model results

Model output inventory with throughput summaries and playable game artifacts.

Frosty Ornith-35B model image

Frosty Ornith-1.0-35B

ModelFrosty Ornith-1.0-35B
Model configGGUF-NVFP4 (qwen35moe) / llama.cpp / Pareto fleet 32
Author: Frosty40 / Ornith
17 games built across 2 prompts
input tok/s2,352.6
decode tok/s277.1
per-agent tok/s8.7
agent Pareto32 · NVIDIA GB10
View results →Download on Hugging Face ↗
AEON QWEN model image

AEON QWEN

ModelAEON QWEN
Model configNVFP4 MTP-XS / vLLM + z-lab DFlash speculative decode / Pareto fleet 8
Author: AEON-7
8 games built across 2 prompts
input tok/s2,057.5
decode tok/s218.3
per-agent tok/s29.9
agent Pareto8 · NVIDIA GB10
View results →Download on Hugging Face ↗
Ornith-AEON model image

Ornith-1.0-35B-AEON-Ultimate-Uncensored-NVFP4

ModelOrnith-1.0-35B-AEON-Ultimate-Uncensored-NVFP4
Model configNVFP4 / vLLM + z-lab DFlash drafter n=6 / team 16 / 20-min build
Author: Ornith-1.0-35B base / AEON-7 config
2 games built across 2 prompts
input tok/s122.7
decode tok/s466.1
per-agent tok/s30.5
agent Pareto16 · NVIDIA GB10
View results →Download on Hugging Face ↗
AgentWorld-35B model image

AgentWorld-35B-A3B

ModelAgentWorld-35B-A3B
Model configGGUF-NVFP4 (qwen35moe) / llama.cpp / Pareto fleet 32
Author: Qwen / AgentWorld
23 games built across 2 prompts
input tok/s1,055.7
decode tok/s145.4
per-agent tok/s4.7
agent Pareto32 · NVIDIA GB10
View results →Download on Hugging Face ↗
Gemma-4-26B model image

Gemma-4-26B-A4B

ModelGemma-4-26B-A4B
Model configNVFP4 / vLLM + DFlash speculative decode / team 32 / 20-min build
Author: NVIDIA / Gemma
2 games built across 2 prompts
input tok/s6,803.0
decode tok/s778.2
per-agent tok/s33.9
agent Pareto32 · NVIDIA GB10
View results →Download on Hugging Face ↗
Gemma-4 current model image

Gemma-4-26B-A4B NVFP4 current plain vLLM

ModelGemma-4-26B-A4B NVFP4 current plain vLLM
Model configNVFP4 / vLLM plain serve, 32k context, max 48 sequences / team 5 / 20-min build
Author: NVIDIA / Gemma
2 games built across 2 prompts
input tok/s298.0
decode tok/s785.9
per-agent tok/s16.4
agent Pareto48 · NVIDIA GB10
View results →Download on Hugging Face ↗
NEX2 model image

Nex-N2-mini

ModelNex-N2-mini
Model configNVFP4 GGUF / llama.cpp / team 3 / 20-min build
Author: nex-agi base / s-batman NVFP4 GGUF conversion
5 games built across 2 prompts
B70 input tok/s32.7
B70 decode tok/s224.3
B70 per-agent tok/s7.9
B70 live fleet32 · Intel Arc Pro B70
View results →Download on Hugging Face ↗
SIQ-1-35B model image

SIQ-1-35B

ModelSIQ-1-35B
Model configQ5_K_M qwen35moe / llama.cpp SYCL on Intel Arc Pro B70 / team 7 / 20-min build
Author: SIQ
4 games built across 2 prompts
B70 input tok/s33.1
B70 decode tok/s223.7
B70 per-agent tok/s7.8
B70 live fleet32 · Intel Arc Pro B70
View results →
Quest Turbo model image

QUEST-30B-RL

ModelQUEST-30B-RL
Model configQ5_K_M qwen3moe / llama.cpp SYCL on Intel Arc Pro B70 / team 6 / 15-min build
Author: Quest / qwen3moe
1 games built across 1 prompts
B70 input tok/s76.3
B70 decode tok/s335.5
B70 per-agent tok/s11.2
B70 live fleet32 · Intel Arc Pro B70
View results →
StepFun Step-3.7 model image

StepFun Step-3.7-Flash

ModelStepFun Step-3.7-Flash
Model configUD-IQ4_NL GGUF / llama.cpp + Q8 MTP draft / team 3 / 20-min build
Author: StepFun
2 games built across 2 prompts
input tok/s94.2
decode tok/s33.6
per-agent tok/s11.3
agent Pareto3 · NVIDIA GB10
View results →
GPT-5.5 xhigh model image

GPT-5.5 xhigh

ModelGPT-5.5 xhigh
Model configfrontier API / Codex subagents / 2 xhigh subagents
Author: OpenAI
2 games built across 2 prompts
View results →
GPT-5.6 Sol model image

GPT-5.6 Sol

ModelGPT-5.6 Sol
Model configfrontier API / Codex Desktop one-shot / independent one-shot output
Author: OpenAI
2 games built across 2 prompts
View results →
GPT-5.6 Terra model image

GPT-5.6 Terra

ModelGPT-5.6 Terra
Model configfrontier API / Codex Desktop one-shot / independent one-shot output
Author: OpenAI
2 games built across 2 prompts
View results →
Claude Opus 4.8 model image

Claude Opus 4.8

ModelClaude Opus 4.8
Model configfrontier API / Claude Code subagents / 2 xhigh subagents
Author: Anthropic
4 games built across 2 prompts
View results →
Claude Sonnet 5 model image

Claude Sonnet 5

ModelClaude Sonnet 5
Model configfrontier API / Claude Code / PR artifact
Author: Anthropic
1 games built across 1 prompts
View results →
DeepSeek V4 Pro model image

DeepSeek V4 Pro

ModelDeepSeek V4 Pro
Model configthinking mode / OpenCode single agent / 1 xhigh subagents
Author: DeepSeek
2 games built across 2 prompts
View results →
GLM-5.2 model image

GLM-5.2

ModelGLM-5.2
Model configthinking mode / OpenCode single agent / 1 xhigh subagents
Author: Zhipu AI
4 games built across 2 prompts
View results →
MoA:AngelX model image

MoA:AngelX

ModelMoA:AngelX
Model configmulti-model orchestration / AngelX MOA improve loop / 9-agent MOA improve loop
Author: AngelX
5 games built across 2 prompts
View results →
Laguna XS 2.1 model image

Laguna XS 2.1

ModelLaguna XS 2.1
Model configQ4_K_M GGUF / llama.cpp/OpenCode on Atlas via local Ollama / single local model run
Author: Poolside / Laguna
3 games built across 2 prompts
B70 input tok/s66.3
B70 decode tok/s283.1
B70 per-agent tok/s10.6
B70 live fleet32 · Intel Arc Pro B70
View results →
MOA:LONGCAT model image

MOA:LONGCAT

ModelMOA:LONGCAT
Model configmulti-model orchestration / Longcat Studios multi-persona MOA / 6-agent MOA improve loop
Author: Longcat Studios
1 games built across 1 prompts
View results →
Qwen3 Coder 30B model image

Qwen3-Coder-30B-A3B-Instruct BF16

ModelQwen3-Coder-30B-A3B-Instruct BF16
Model configBF16 safetensors / vLLM BF16, 131k context for game builds / team 6 / 20-min build
Author: Qwen
2 games built across 2 prompts
input tok/s87.0
decode tok/s344.2
per-agent tok/s10.8
agent Pareto32 · NVIDIA GB10
View results →Download on Hugging Face ↗
Nemotron 3 Super model image

NVIDIA Nemotron-3-Super-120B-A12B-NVFP4

ModelNVIDIA Nemotron-3-Super-120B-A12B-NVFP4
Model configNVFP4 safetensors / vLLM 0.23.0, 8k context, max_num_seqs=1 / team 1 / 20-min build
Author: NVIDIA
2 games built across 2 prompts
input tok/s4.9
decode tok/s13.9
per-agent tok/s13.9
agent Pareto1 · NVIDIA GB10
View results →Download on Hugging Face ↗
Nemotron Puzzle 75B model image

NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-NVFP4

ModelNVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-NVFP4
Model configNVFP4 + FP8 ModelOpt safetensors / vLLM 0.23.0, 8k context, MTP n=1, max_num_seqs=32 / team 1 / 20-min build
Author: NVIDIA
2 games built across 2 prompts
input tok/s114.4
decode tok/s319.8
per-agent tok/s10.3
agent Pareto32 · NVIDIA GB10
View results →Download on Hugging Face ↗
Qwen3.6-35B Base model image

Qwen3.6-35B-A3B UD-Q6_K GGUF

ModelQwen3.6-35B-A3B UD-Q6_K GGUF
Model configUD-Q6_K GGUF / llama.cpp-qwen36, 98k context, q8 KV / team 6 / 20-min build
Author: Qwen
1 games built across 1 prompts
input tok/s64.7
decode tok/s244.6
per-agent tok/s7.6
agent Pareto32 · NVIDIA GB10
View results →
Qwen3.6-35B Q5 Turbo model image

Qwen3.6-35B-A3B UD-Q5_K_XL GGUF

ModelQwen3.6-35B-A3B UD-Q5_K_XL GGUF
Model configUD-Q5_K_XL GGUF / llama.cpp SYCL on Intel Arc Pro B70, 12-slot unified KV / Pareto fleet 12
Author: Qwen / local Turbo
2 games built across 2 prompts
input tok/s61.8
decode tok/s158.5
per-agent tok/s16.3
agent Pareto12 · Intel Arc Pro B70
View results →
Qwen3.6-35B NVFP4 model image

nvidia/Qwen3.6-35B-A3B-NVFP4

Modelnvidia/Qwen3.6-35B-A3B-NVFP4
Model configNVFP4 (W4A16) + FP8 attn / vLLM 0.23.0, modelopt NVFP4+FP8, MARLIN MoE / team 32 / 20-min build
Author: NVIDIA
2 games built across 2 prompts
input tok/s174.6
decode tok/s660.3
per-agent tok/s20.6
agent Pareto32 · NVIDIA GB10
View results →Download on Hugging Face ↗
Leanstral NVFP4 model image

Leanstral-1.5-119B-A6B GGUF-NVFP4

ModelLeanstral-1.5-119B-A6B GGUF-NVFP4
Model configGGUF-NVFP4 / llama.cpp local OpenAI-compatible server / single local model run
Author: Mistral AI base / Frosty40 GGUF-NVFP4 conversion
2 games built across 2 prompts
View results →Download on Hugging Face ↗
LongCat-2.0 model image

Meituan LongCat-2.0 (1.6T MoE, 1M ctx)

ModelMeituan LongCat-2.0 (1.6T MoE, 1M ctx)
Model configFP8 (Meituan) / Hosted API (api.longcat.chat), OpenAI-compatible / hosted API one-shot + chain3
Author: Meituan
8 games built across 2 prompts
View results →Download on Hugging Face ↗
Fable model image

Fable

ModelFable
Model configexternal submission / submitted single-file HTML artifacts / submitted artifact
Author: Fable
2 games built across 2 prompts
View results →
Agents-A1 model image

InternScience/Agents-A1 (35B MoE, agentic)

ModelInternScience/Agents-A1 (35B MoE, agentic)
Model configBF16 (unquantized) / vLLM 0.23.0, BF16, FlashInfer CUTLASS MoE / team 32 / 20-min build
Author: InternScience
2 games built across 2 prompts
input tok/s92.2
decode tok/s348.9
per-agent tok/s10.9
agent Pareto32 · NVIDIA GB10
View results →Download on Hugging Face ↗
UNDER RAPID CONSTRUCTION - WILL CHANGE
Frostmark retro console footer
why

Local agents need public test loops

Working on local agents matters because these tests are the first step toward an R&D base-model trust score: a practical signal for choosing which architectures and models deserve deeper research and development.

Agentic Arcade is a playable hello world for that work. Anyone can open the games, review the artifacts, and compare model behavior. The mission is maximal distribution of intelligence to all people.