// AI Platform Drift
DeepSeek V4 Flash completion price cut 25%, prompt price up 12.5%
OpenRouter model catalogue · published 2026-10-13 00:42 UTC
CORRECTED 2026-10-14 02:35 UTC — CORRECTION: The thresholds of 53 completion tokens per prompt token in WHY IT MATTERS and 53:1 in WHAT TO DO are wrong. Using only the captured base rates for ~deepseek/deepseek-v4-flash-latest, break-even is exactly 1 completion token(s) per 120 prompt token(s); total request cost is lower above that completion-to-prompt ratio, equal at that ratio, and higher below it, excluding all other charges. This is one output token per 120 input tokens (output/input = 1/120, approximately 0.008333); equality is break-even, not savings. At ten completion tokens per prompt token, total request cost falls 24.9%, consistent with WHAT CHANGED. Use that corrected threshold when interpreting the original budget advice. Captured arithmetic only; provider rate validity, applicability, event date and effective conditions remain unverified.
Stood 25h 52m · cause arithmetic · every correction ↗
WHAT CHANGEDOpenRouter's pricing for ~deepseek/deepseek-v4-flash-latest changed on 13 October 2026. The captured completion base rate changed from $0.768 to $0.576 per 1M tokens. The captured prompt base rate changed from $0.0128 to $0.0144 per 1M tokens. For workloads generating ten completion tokens for every prompt token, total cost per request decreased 24.9%.
WHY IT MATTERSCompletion-heavy workloads cost less; prompt-heavy workloads cost more. Using the captured base rates for ~deepseek/deepseek-v4-flash-latest, 100M completion tokens per day cost $76.80 before and $57.60 after, excluding all other charges. Using the captured base rates for ~deepseek/deepseek-v4-flash-latest, 100M prompt tokens per day cost $1.28 before and $1.44 after, excluding all other charges. If your application generates more than 53 completion tokens per prompt token, you save money; below that ratio, you pay more.
WHAT TO DOPull last month's token usage by type from your logs or OpenRouter dashboard. Calculate your actual completion-to-prompt ratio. If you are below 53:1, reforecast your monthly spend using the new rates and update your budget. If you are above 53:1, document the savings for your next cost review.
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: 0.0000000128 -pricing.completion: 0.000000768 +pricing.prompt: 0.0000000144 +pricing.completion: 0.000000576 context_length: 1048576 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 for an active model directly affect engineering cost calculations and budget planning. |
| signals | pricing.prompt increased: 0.0000000128 → 0.0000000144 USD/token · pricing.completion decreased: 0.000000768 → 0.000000576 USD/token · model pricing change on active catalogue entry |
| suppression | — |
| triage model | anthropic/claude-haiku-4.5 ($0.00000) |
| writer model | anthropic/claude-sonnet-4.5 ($0.00000) |