Benchmarks · Writing & design · Games & strategy
GameFeel Reverse Evaluation v1.5
How well a model gives feedback on proposed changes to unpublished games, especially on balance. GameFeel's reverse evaluation, made much harder in v1.5: two thirds runs with no help, one third runs where the model first sees 39 worked examples. Out of 100.
26 models · 33 results · updated 2026-11-13 · kodasdream.github.io ↗
Leaderboard
Fable 5.2 high
More from this benchmark
2 published measuresScore vs any published measure2 measures: sub-scores, splits, counts
13 more models under All models in settings.
Breakdown
darker is better within each column · click a column to sort| Model | ||
|---|---|---|
| Fable 5.2 high | 22.8 | 88.1 |
| Gemini 4 Carbon high | 32.8 | 54.5 |
| Gemini 4 Hydrogen high | 7.55 | 56.6 |
| Opus 5.5 xhigh | 13.4 | |
| Sonnet 5.5 max | 9.82 | |
| Gemini 4 Argon high | 6.69 | |
| Qwen 3.8 Flash xhigh | 6.46 | |
| Qwen 3.8 Max xhigh | 6.2 | |
| Qwen 3.7 Plus medium | 4.01 | |
| GPT-6.1 Sol xhigh | 3.93 | |
| GPT-6.1 Terra xhigh | 3.71 | |
| DeepSeek V4.1 Flash 1017 max | 2.85 | |
| MiMo V2.6 Pro max | 2.1 | |
| DeepSeek V4.1 Flash max | 1.7 | |
| Haiku 5.5 max | 1.53 | |
| Ling 3.2 Flash max | 1.5 | |
| Gemini 3.8 Flash max | 0.93 | |
| Nemotron 3 Ultra max | 0.89 | |
| Muse Wonder max | 0.48 | |
| MiniMax M3.1 max | 0.43 | |
| Muse Spark 1.3 medium | 0.39 | |
| GLM 5.3 Flash max | 0.36 | |
| Kimi K3.1 max | 0.1 | |
| GPT-6 Luna xhigh | 0 | 0.26 |
| Qwen 3.8 27B xhigh | 0.02 |
All results
33 rows| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Fable 5.2 | high | 44.1 |
| 2 | Gemini 4 Carbon | high | 39.6 |
| 3 | Gemini 4 Carbon | medium | 37.8 |
| 4 | Gemini 4 Carbon | low | 25.0 |
| 5 | Gemini 4 Hydrogen | high | 23.7 |
| 6 | Fable 5.2 | medium | 22.7 |
| 7 | Fable 5.2 | max | 20.0 |
| 8 | Opus 5.5 | xhigh | 13.2 |
| 9 | Sonnet 5.5 | max | 9.7 |
| 10 | Opus 5.5 | medium | 7.5 |
| 11 | Sonnet 5.5 | xhigh | 7.0 |
| 12 | Gemini 4 Argon | high | 6.6 |
| 13 | Qwen 3.8 Flash | xhigh | 6.4 |
| 14 | Qwen 3.8 Max | xhigh | 6.1 |
| 15 | Sonnet 5.5 | medium | 4.6 |
| 16 | Qwen 3.7 Plus | medium | 4.0 |
| 17 | GPT-6.1 Sol | xhigh | 3.9 |
| 18 | GPT-6.1 Terra | xhigh | 3.7 |
| 19 | DeepSeek V4.1 Flash 1017 | max | 2.8 |
| 20 | MiMo V2.6 Pro | max | 2.1 |
| 21 | DeepSeek V4.1 Flash | max | 1.7 |
| 22 | Haiku 5.5 | max | 1.5 |
| 23 | Ling 3.2 Flash | max | 1.5 |
| 24 | Gemini 3.8 Flash | max | 0.9 |
| 25 | Nemotron 3 Ultra | max | 0.9 |
| 26 | Muse Wonder | max | 0.5 |
| 27 | MiniMax M3.1 | max | 0.4 |
| 28 | Muse Spark 1.3 | medium | 0.4 |
| 29 | GLM 5.3 Flash | max | 0.4 |
| 30 | Kimi K3.1 | max | 0.1 |