DeepSeek to Quadruple API Prices for V4 Pro and V4 Flash Starting August 16
DeepSeek will raise API output token pricing roughly fourfold starting August 16, introducing peak and off-peak rates for its V4 Pro and V4 Flash models. Despite the increase, DeepSeek remains cheaper than competitors like OpenAI's GPT-5.6 Sol and Moonshot's Kimi K3.
DeepSeek is raising API pricing for its V4 model lineup by roughly four times starting August 16, ending the aggressive discount strategy that helped the Chinese AI lab undercut Western competitors.
According to the company, DeepSeek V4 Pro will cost $3.96 per 1 million output tokens during peak hours, up from the current $0.87 — a 4.5x increase. Off-peak pricing will run at $1.98 per 1 million output tokens, half the peak rate.
The smaller DeepSeek V4 Flash model will see output pricing rise from $0.28 to $1.32 per 1 million tokens at peak hours, with off-peak pricing set at $0.66.
DeepSeek said the new peak and off-peak structure is designed to "allocate resources more reasonably," suggesting the tiered pricing is meant to shift compute load away from high-demand hours rather than purely to increase revenue.
From promotion to permanent hike
The pricing shift reverses an earlier commitment. The discounted rates customers are currently paying were originally billed as a promotion set to expire May 31. In May, DeepSeek said it planned to make those discounted prices permanent. Instead, the company is now moving in the opposite direction, tying the increase to the rollout of its latest model, DeepSeek V4 Pro.
Still cheaper than rivals
Even after the fourfold increase, DeepSeek's pricing remains well below several competitors on a per-token basis. Moonshot AI's Kimi K3 costs $15 per 1 million output tokens — roughly 3.8x DeepSeek V4 Pro's new peak rate. OpenAI's flagship model, GPT-5.6 Sol, costs $30 per 1 million output tokens, nearly 8x DeepSeek's peak pricing.
OpenAI's lower-cost option, GPT-5.6 Luna, priced at $1.20 per 1 million output tokens, actually undercuts DeepSeek V4 Flash's new peak rate of $1.32, though it remains more expensive than DeepSeek's off-peak Flash pricing of $0.66.
What this means
DeepSeek's original disruption of the AI market rested heavily on radically undercutting Western labs on price while offering comparable performance. This increase — while still leaving DeepSeek cheaper than GPT-5.6 Sol and Kimi K3 — signals that the company's cost structure or business priorities have shifted enough to abandon a pledge to keep discounted rates permanent.
The peak/off-peak split is notable beyond the price hike itself: it points to real compute capacity constraints, forcing DeepSeek to use pricing as a demand-shaping tool rather than relying purely on flat-rate simplicity. Enterprise customers running high-volume workloads may now need to actively schedule inference around off-peak windows to capture savings, adding operational complexity that wasn't previously necessary.
For the broader market, the move suggests the era of DeepSeek functioning as an unambiguous low-cost anchor may be ending, even if it still undercuts most frontier-model competitors by a wide margin.
Related Articles
DeepSeek removes separate fees for model-triggered search and web fetch from Nov 18
DeepSeek removed separate charges for server-side search and web fetch triggered during model calls, effective at noon on November 18, 2026, Beijing time. Token usage is still billed, and the standalone Search and Fetch APIs remain at 5 yuan per 1,000 successful calls.
DeepSeek opens Search, Fetch, OCR, ASR and TTS as standalone APIs, with tool fees billed on top of tokens
DeepSeek has made the capabilities behind its app available as standalone APIs: Search and Fetch, OCR, ASR and TTS. It also open-sourced chat-demo, a static web client that connects to the API from the browser. Tool use is billed on top of tokens.
OpenAI tells developers to rewrite system prompts for GPT-6.2 Astra and Sol as defaults shift
OpenAI says developers moving to GPT-6.2 Astra or Sol should review their system prompts. The company published copy-ready templates in the models' documentation, and the Responses API now discards earlier reasoning content if conversation history is edited.
OpenAI says GPT-6.2 Astra safeguards cut reasoning-extraction attack success to 0.2%
OpenAI announced on November 20, 2026 stronger defenses against adversarial distillation. It reports that adaptive attacks extracting protected reasoning from an earlier request succeed 0.2% of the time on GPT-6.2 Astra with updated safeguards, versus 54.0% on GPT-6 Astra with original safeguards.
Comments
Loading...