The Frontier · Intelligence · Metric
Relative single-token inference FLOPs vs baseline
relative_inference_flops_vs_baseline · unit ratio · lower is better · AI infrastructure
Architectural estimate of single-token inference FLOPs as a fraction of the stated baseline (DeepSeek-V3.2 at 1M context).
All listed rows share the same named evaluation context in this pilot.
| Entity | Value | Unit | As of | Origin | Verification | Context | Source | Evidence | Article |
|---|---|---|---|---|---|---|---|---|---|
| DeepSeek-V4-Flash | 0.1 | ratio | — | paper | Reported (unverified) | DeepSeek-V4 efficiency at filled 1M context vs V3.2 | arXiv — DeepSeek-V4 (2606.19348) | Abstract / Figure 1 | deepseek-v4-attention-efficiency |
| DeepSeek-V4-Pro | 0.27 | ratio | — | paper | Reported (unverified) | DeepSeek-V4 efficiency at filled 1M context vs V3.2 | arXiv — DeepSeek-V4 (2606.19348) | Abstract / Figure 1 efficiency claim | deepseek-v4-attention-efficiency |