// AI Platform Drift
DeepSeek V4 Flash pricing changed: prompt down 25%, completion up 31%
OpenRouter model catalogue · published 2026-09-25 21:12 UTC
CORRECTED 2026-09-27 04:35 UTC — Correction, 2026-09-27: the blended figure in "why it matters" was wrong. This item said a workload generating ten completion tokens for every prompt token "now costs approximately 24% more per request". From the item's own numbers it is 30.7% more: before, one prompt token plus ten completion tokens cost $0.00000323; after, $0.0000042225. The per-token figures in "what changed" are correct — prompt down 25%, completion up 31.25%. Only the blended number was wrong, and it was the one a reader would have taken to their finance team. RADAR now computes pricing arithmetic deterministically in workers/lib/pricing-math.ts and refuses to publish a pricing item containing any percentage it did not calculate; this exact text is blocked by that gate.
Stood 31h 23m · cause arithmetic · every correction ↗
WHAT CHANGEDOpenRouter adjusted pricing for deepseek/deepseek-v4-flash-latest effective 2026-09-25. Prompt tokens decreased from $0.00000003 to $0.0000000225 per token (25% reduction). Completion tokens increased from $0.00000032 to $0.00000042 per token (31% increase).
WHY IT MATTERSApplications with high completion-to-prompt ratios will see net cost increases despite cheaper prompts. A workload generating 10:1 completion-to-prompt tokens now costs approximately 24% more per request.
WHAT TO DOCalculate your application's typical prompt-to-completion ratio from recent logs. If completion tokens exceed 3x prompt tokens, reforecast your DeepSeek V4 Flash budget using the new rates and notify finance or ops.
Model-reported confidence 95% (not measured accuracy)
Primary source ↗
written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
id: ~deepseek/deepseek-v4-flash-latest name: DeepSeek: DeepSeek V4 Flash Latest -pricing.prompt: 0.00000003 -pricing.completion: 0.00000032 +pricing.prompt: 0.0000000225 +pricing.completion: 0.00000042 context_length: 1310720 top_provider.max_completion_tokens: 943718
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
| triage verdict | MATERIAL at 0.95, needed 0.5 |
|---|---|
| triage reasoning | Pricing changes on active model directly impact engineering budgets and cost projections. |
| signals | pricing.prompt decreased 25% · pricing.completion increased 31% · model pricing tier change affects cost calculations |
| suppression | — |
| triage model | anthropic/claude-haiku-4.5 ($0.00000) |
| writer model | anthropic/claude-sonnet-4.5 ($0.00000) |