// AI Platform Drift
DeepSeek V4 Flash 0731 completion price halved, prompt price up 69.5%
OpenRouter model catalogue · published 2026-10-11 05:48 UTC
CORRECTED 2026-10-18 23:27 UTC — [CORRECTION · WHAT TO DO] Original: “If you are prompt-heavy (more than three prompt tokens per completion token), verify the net increase fits your budget.” Fact: At the captured base rates, total cost rises only above about 97 prompt tokens per completion token (0.64 / 0.0066); at 3 prompt tokens per completion token, total cost falls about 47%. / [CORRECTION · WHY IT MATTERS] Original: “Prompt-heavy workloads with minimal generation still benefit but less dramatically.” Fact: Above about 97 prompt tokens per completion token total cost rises; a prompt-only workload pays 69.5% more. / Captured arithmetic only; vendor rate validity, applicability and effective conditions remain unverified.
Stood 185h 38m · cause arithmetic · every correction ↗
WHAT CHANGEDOpenRouter's pricing for deepseek/deepseek-v4-flash-0731 changed on 2026-10-11. The captured completion base rate changed from $1.28 to $0.64 per 1M tokens. The captured prompt base rate changed from $0.0095 to $0.0161 per 1M tokens.
WHY IT MATTERSThe net effect depends on your completion-to-prompt ratio. Workloads generating ten completion tokens for every prompt token see a 49.9% total cost reduction. Workloads with equal prompt and completion volumes see a 49.1% reduction. Prompt-heavy workloads still pay less unless they send more than about 97 prompt tokens per completion token; a prompt-only workload pays 69.5% more. Using the captured base rates for deepseek/deepseek-v4-flash-0731, 100M completion tokens per day cost $128.00 before and $64.00 after, excluding all other charges.
WHAT TO DOPull your last 30 days of token usage for this model from OpenRouter's API, calculate your actual prompt-to-completion ratio, and reforecast your monthly spend using the new rates. Total cost rises only above about 97 prompt tokens per completion token (at 3 prompt tokens per completion token it falls about 47%); if your workload is above that ratio, verify the increase fits your budget.
Original text (as first published, superseded)
WHY IT MATTERSThe net effect depends on your completion-to-prompt ratio. Workloads generating ten completion tokens for every prompt token see a 49.9% total cost reduction. Workloads with equal prompt and completion volumes see a 49.1% reduction. Prompt-heavy workloads with minimal generation still benefit but less dramatically. Using the captured base rates for deepseek/deepseek-v4-flash-0731, 100M completion tokens per day cost $128.00 before and $64.00 after, excluding all other charges.
WHAT TO DOPull your last 30 days of token usage for this model from OpenRouter's API, calculate your actual prompt-to-completion ratio, and reforecast your monthly spend using the new rates. If you are prompt-heavy (more than three prompt tokens per completion token), verify the net increase fits your budget.
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-0731 name: DeepSeek: DeepSeek V4 Flash 0731 -pricing.prompt: 0.0000000095 -pricing.completion: 0.00000128 +pricing.prompt: 0.0000000161 +pricing.completion: 0.00000064 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 change on active model affects cost calculations for existing workloads. |
| signals | pricing.prompt increased 69.5% (0.0000000095 → 0.0000000161 USD per token) · pricing.completion decreased 50% (0.00000128 → 0.00000064 USD per token) · asymmetric price movement across token types |
| suppression | — |
| triage model | anthropic/claude-haiku-4.5 ($0.00000) |
| writer model | anthropic/claude-sonnet-4.5 ($0.00000) |