DeepSeek V4 Pro 0813 API Prices Double: Input Hits $1.32, Output $3.96 per 1M Tokens

What Changed

DeepSeek quietly pushed a +100% price increase on its V4 Pro 0813 API. Input tokens now cost $1.32 per 1M (up from $0.66) and output tokens $3.96 per 1M (up from $1.98). Both figures doubled in lockstep — no tiered rollout, no grandfathering window announced.

See the full timeline on the DeepSeek V4 Pro 0813 — live specs & price history page.

Does It Matter for Your Workload?

A doubling is material regardless of starting point, but context matters:

• High-output tasks (long-form generation, coding assistants, RAG pipelines) feel this hardest. At $3.96/1M output tokens, a workflow burning 10M output tokens/month jumps from ~$19.80 to ~$39.60 — a real budget line.
• Read-heavy / short-output tasks (classification, embeddings proxies, short summarisation) will see a more modest absolute hit, since input at $1.32/1M is still competitive in the mid-range tier.
• Batch or eval workloads running millions of tokens for testing are where this bites most unexpectedly. Re-run your cost models before your next billing cycle.

What to Do Now

1. Audit token consumption split by input vs. output for the last 30 days.
2. Compare alternatives on AIToolScout — several models in the same capability band currently sit below $1.32 input / $3.96 output.
3. Lock in any cached pricing if DeepSeek's platform offers prompt caching discounts; they can offset output-heavy costs significantly.

Price increases of this magnitude from a previously budget-friendly provider are worth treating as a procurement event, not just a footnote.

Ezra, Scout AI Team