// AI Platform Drift
DeepSeek Flash Latest pricing and output limits changed
OpenRouter model catalogue · published 2026-09-29 04:41 UTC
CORRECTED 2026-10-02 17:58 UTC — [CORRECTION · WHAT CHANGED] Original: “completion pricing rose from $384/M to $943.72/M (+145.8%)” Fact: Completion pricing rises from $0.39/M to $0.60/M (+53.8%). Those other values originate from output-token limits, not price fields.
[CORRECTION · WHY IT MATTERS] Original: “Total cost per request increases 145.8% across all workload mixes because completion tokens now cost more than twice as much, outweighing the prompt discount. Applications generating long outputs will see the largest absolute cost increase, though the higher token limit enables responses that were previously truncated.” Fact: Net cost depends on the token mix: prompt -66.7%, completion +53.8%. Do not infer a universal increase from token-limit growth. Higher limits are capacity, not dollar rates.
[CORRECTION · WHAT TO DO] Original: “$943.72/M completion” Fact: Use the captured completion rate of $0.60/M alongside $0.02/M prompt, subject to applicable source conditions.
Stood 85h 17m · cause extraction error · every correction ↗
Captured arithmetic only; vendor rate validity, applicability and effective conditions remain unverified.
WHAT CHANGEDOpenRouter updated deepseek/deepseek-flash-latest on 2026-09-29: prompt pricing dropped from $0.06/M to $0.02/M (-66.7%), completion pricing rose from $384/M to $943.72/M (+145.8%), and max_completion_tokens increased from 384,000 to 943,718 (+146%).
WHY IT MATTERSTotal cost per request increases 145.8% across all workload mixes because completion tokens now cost more than twice as much, outweighing the prompt discount. Applications generating long outputs will see the largest absolute cost increase, though the higher token limit enables responses that were previously truncated.
WHAT TO DOPull current month-to-date request logs for this model, calculate new costs using $0.02/M prompt and $943.72/M completion, and update your budget forecast. If cost increase is material, evaluate whether the expanded output limit justifies it or whether to switch models.
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-flash-latest name: DeepSeek: DeepSeek Flash Latest -pricing.prompt: 0.00000006 -pricing.completion: 0.00000039 +pricing.prompt: 0.00000002 +pricing.completion: 0.0000006 context_length: 1048576 -top_provider.max_completion_tokens: 384000 +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 0.98, needed 0.5 |
|---|---|
| triage reasoning | Pricing and token limits changed significantly; teams budgeting or optimizing this model need immediate notice. |
| signals | prompt pricing decreased 66% · completion pricing increased 54% · max_completion_tokens increased 146% |
| suppression | — |
| triage model | anthropic/claude-haiku-4.5 ($0.00000) |
| writer model | anthropic/claude-sonnet-4.5 ($0.00000) |