Model results

GraphWalks

Choose a shared evaluation protocol to compare model scores. Versions, sources and configurations stay separate.

Atlas snapshot captured 10/19/2026, 10:20:28 AM UTC. This is an import timestamp, not the evaluation date. Coverage may be incomplete.

Official source

← All evaluation groups

GraphWalks

Source: Google DeepMind model evaluation vendor-reported (reviewed publisher results)
Version: deepmind-argon-october-2026-dataset-revision-unreported
Measurement date: Not reported
Metric: % F1 · Higher is better
Evaluation scope: BFS F1, 256k to 1M tokens, 200 items; all four models evaluated by Google DeepMind for the Gemini 4 Argon October 2026 evaluation report, pages 3 and 5. Gemini uses the API at highest thinking settings; non-Gemini models use maximum available thinking/reasoning where available, otherwise best available reasoning. Exact per-model effort and dated endpoint revision are not specified. Model mapping is release-level only. Dataset revision and whether the February 2026 prompt/label corrections were used are not reported. This is F1, not perfect success or the original GPT-4.1 launch accuracy. result date not supplied; captured 2026-10-19.

Positions are withheld because at least one result lacks explicit independent provenance or a documented evaluation scope. Missing metadata is shown as unreported; matching imported metadata is not independent verification of the protocol.

Gemini 4 Argon
google/gemini-4-argon-20261013
84.2% F1
View modelEvidence model-card-reported
GPT-6 Astra
openai/gpt-6-astra-20260903
71.8% F1
View modelEvidence model-card-reported
Claude Opus 5.5
anthropic/claude-opus-5-5
66.8% F1
View modelEvidence model-card-reported
Claude Fable 5.1
anthropic/claude-fable-5-1
65% F1
View modelEvidence model-card-reported