Anthropic / MODEL EVIDENCE
1 benchmarks · 2 observations · 1 sources
1 benchmarks · large dot: DeepSeek V4.1 Flash / Claude Code; small dots: closest models where reported
DeepSeek V4.1 Flash / Claude Code · xdev-Bench-hard · pass@3 · Kimi K3.1 Pro · internal xdev-bench-hard ↗
57.1% · DeepSeek V4.1 Flash / Claude Code · xdev-Bench-hard · Average success rate · Kimi K3.1 Pro · internal xdev-bench-hard · Introducing Kimi K3.1 Pro ↗
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
1 of 1 shared benchmark comparisons
1 of 1 shared benchmark comparisons
1 of 1 shared benchmark comparisons
Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.
POST-TRAINING / SOURCE DISCLOSURE
The Flash technical report describes a supervised starting stage, reinforcement learning across task domains, then distillation from an ensemble of teachers. It emphasizes synthesized tasks and scalable interactive environments.
The report documents Flash. The 1017 card calls that release a minor update but does not supply a separate complete recipe; these stages should not be assumed to describe every DeepSeek V4.1 variant.
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
DeepSeek V4.1 Flash / Claude Code