Skip to content

Benchmark

GraphWalks BFS

Scores are percentages. Higher is better.

Models measured
14
Scores recorded
42
Sources
1
IQ weight
0.19

Score distribution

71.4%
Mean
75%
Median
16.1%
Std dev
45.4%
10th percentile
86%
90th percentile
21.4%
Min
21.4%94%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.19 for chat models, measured over 14 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1GPT 5.2<128K · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.94%
2GPT 5.20K-128K · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.94%
3GPT 5.40K-128K · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.93%
4GPT-5.6 Sol256K · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.90.7%
5MiMo V2.5official model-card chart · F1 · 32K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.86%
6Claude Opus 4.8256K · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.9%
7Claude Opus 4.8Fable/Mythos 5 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · 256K · F1· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.9%
8Claude Opus 4.8Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · 256K · F1· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85.9%
9MiMo V2.5official model-card chart · F1 · 64K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.85%
10MiMo V2.5 Proofficial model-card chart · F1 · 128K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82%
11GPT-5.6 Luna256K · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.3%
12MiMo V2.5 Proofficial model-card chart · F1 · 32K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81%
13MiMo V2.5 Proofficial model-card chart · F1 · 64K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79%
14MiMo V2.5official model-card chart · F1 · 256K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78%
15GPT-5.6 Sol1M · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.77.1%
16MiMo V2.5 Proofficial model-card chart · F1 · 256K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.77%
17Claude Opus 4.7public-API-reproducible 256K subset · corrected F1 · avg@5· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.9%
18Claude Opus 4.7Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · 256K · F1· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.9%
19GPT-5.6 Terra256K · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.9%
20GPT-5.4 minixhigh · 0K–128K· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.76.3%
21MiMo V2.5official model-card chart · F1 · 128K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.75%
22Claude Sonnet 4.6256K API-reproducible subset of 1M · F1 · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.74.5%
23Claude Sonnet 4.61M set · F1 · adaptive thinking · max effort · 5 trials· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.8%
24GPT 5.5256K · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.7%
25GPT 5.5256K · F1 · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.7%
26GPT 5.5Fable/Mythos 5 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · 256K · F1· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.7%
27GPT 5.5Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted · 256K · F1· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.7%
28GPT 5.4 Nanoxhigh · 0K–128K· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.4%
29Claude Sonnet 4.6256K API-reproducible subset of 1M · F1 · 64K extended thinking· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.72.8%
30GPT-5.6 Terra1M · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.71.2%
31Claude Sonnet 4.61M set · F1 · 64K extended thinking · 5 trials· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.68.4%
32Claude Opus 4.81M · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.68.1%
33Claude Opus 4.7256K–1M aggregate · corrected F1 · avg@5· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.58.6%
34MiMo V2.5official model-card chart · F1 · 512K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.58%
35MiMo V2.5 Proofficial model-card chart · F1 · 512K input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.56%
36MiMo V2.5official model-card chart · F1 · 1M input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.54%
37GPT-5.6 Luna1M · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51.2%
38GPT 5.51M · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.45.4%
39GPT 5.51M · F1 · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.45.4%
40Claude Opus 4.6Anthropic Opus 4.6 launch page · adaptive thinking · max effort unless noted · 1M · BFS · F1· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.38.7%
41MiMo V2.5 Proofficial model-card chart · F1 · 1M input tokens· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.37%
42GPT 5.4256K-1M · F1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.21.4%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.