Model ledger

DeepSeek V4 Flash

OpenCode Go · Provider not separately recorded · opencode-go · 24 recorded runs · 18 benchmarks. Open a run for its public headline result. Paid research access opens the full available ledger and configuration. Builder adds exact prompts, released projects, the RemakeBench Harness and production skills for the tests in its library.

Max reasoningVision exp reasoning
Benchmark

Campfire under the stars

2 runs · Max, Vision exp reasoning

Benchmark

Courtyard Blossom Stage 2

1 run · Vision exp reasoning

Benchmark

ECHO Main Menu

1 run · Vision exp reasoning

Benchmark

Explorable space-flight game

1 run · Max reasoning

Benchmark

Fighter Game v1

1 run · Max reasoning

Benchmark

High-voxel-density Jungle Temple diorama

1 run · Max reasoning

Benchmark

High-voxel-density Oasis Outpost diorama

1 run · Vision exp reasoning

Benchmark

High-voxel-density Shrine Village diorama

1 run · Max reasoning

Benchmark

Infinite cathedral corridor shader

2 runs · Max, Vision exp reasoning

Benchmark

Interactive Three.js voxel map viewer

2 runs · Max, Vision exp reasoning

Benchmark

JRPG boss battle

2 runs · Max, Vision exp reasoning

Benchmark

MacBook-class cinematic ad scene

1 run · Max reasoning

Benchmark

Neo-gothic storm city shader

2 runs · Max, Vision exp reasoning

Benchmark

Neon Gunner

2 runs · Max, Vision exp reasoning

Benchmark

Off-road Mud Game

1 run · Max reasoning

Benchmark

Sekiro Main Menu

1 run · Vision exp reasoning

Benchmark

Space Flight Stage 2 — Landfall

1 run · Max reasoning

Benchmark

The Last of Us Main Menu

1 run · Vision exp reasoning

RemakeBenchResearch console