DeepSeek V4 Pro: Official GA Release & New Peak/Off-Peak API Pricing
DeepSeek officially released V4 Pro (GA) on August 13, 2026, and introduced a new peak/off-peak pricing model for its API effective 16:00 UTC on August 16, 2026.
What's new
🟢 OpenAI + Anthropic compatible — same base_url, just swap the model value (deepseek-v4-pro or deepseek-v4-flash)
🟢 1M token context window — handles full codebases, long documents, multi-step agents
🟢 Better JSON reliability — JSON parsing rate up from 78% to 85% (97% with regex guidance)
🟢 Open weights available — self-hosting option for full data privacy
New pricing (per 1M tokens, from Aug 16)
V4 Flash — off-peak: 🟢 Input (cache hit): $0.007 🟢 Input (cache miss): $0.22 🟢 Output: $0.66
V4 Flash — peak (2x off-peak): $0.014 / $0.44 / $1.32
V4 Pro — off-peak: 🟢 Input (cache hit): $0.022 🟢 Input (cache miss): $0.66 🟢 Output: $1.98
V4 Pro — peak (2x off-peak): $0.044 / $1.32 / $3.96
Peak hours are 01:00–04:00 and 06:00–10:00 UTC (4 hours/day); the rest is off-peak at half the price.
How it compares
🟢 DeepSeek V4 Pro (off-peak): $0.66 / $1.98 per M tokens 🟡 Claude Fable 5: $10 / $50 — output ~25x more expensive 🟡 GPT-5.5: $5 / $30 🟡 Claude Opus 5: $5 / $25
DeepSeek V4 Pro scores 62.7 on DeepSWE v1.1 (ahead of Claude Opus 4.8 at 58.0), making it one of the strongest value picks for coding and agents.
Bottom line
For UAE businesses, the new pricing makes AI projects — WhatsApp bots, customer-service agents, document processing — economically viable even for small companies. API compatibility with OpenAI and Anthropic means you can switch models without rewriting your stack.
📚 DeepSeek Harness — open-source agent framework 📚 GPT-5.6 API guide — pricing and usage 📚 Claude Fable 5 — complete guide
Sources
- DeepSeek API — official pricing page (peak/off-peak)
- DeepSeek-V4-Pro GA announcement
- DeepSeek API changelog
How Katbi Can Help
🤖 Custom AI Agents — autonomous task execution for your business 📱 AI WhatsApp Bots — 24/7 customer service on the best cost-performance models 🌐 Web & Mobile Apps — with custom AI integration 💰 API cost optimization — we pick the right model and timing to cut your bills
Related Articles
GLM-5.3 by Z.ai: The Best Open-Source Coding Model of 2026 — Complete Guide
GLM-5.3 from Z.ai is the most capable open-weights coding model — 50% better than GLM-5.2, state-of-…
Claude Managed Agents: The End of Building Your Own Agent Infrastructure
Anthropic just shipped a fully managed agent runtime — no more custom loops, sandboxes, or tool laye…
How We Use AI to Deliver Better Results for Our Clients
Real examples of how AI transformed our workflow — from faster development to smarter solutions and …