DeepSeek V4 Fixed Pricing and Automatic V4-Pro-0813 Upgrade

Date: 2026-08-14 / (1405-05-23)

Update — September 11, 2026: From September 14, 2026 at 04:00 UTC, AvalAI will route deepseek-v4-pro to deepseek-v4.1-flash and bill at V4.1 Flash rates: $0.15 input, $0.003 cached input, and $0.60 output per 1M tokens, fixed at all hours. Use deepseek-v4.1-flash now and test before the cutover. The V4-Pro upgrade, prices, and recommendations below are historical and describe the period before that change. See the new model announcement.

Summary

DeepSeek is introducing peak and off-peak API pricing on August 16, 2026. AvalAI will provide the lower off-peak DeepSeek V4 rates at all hours and has automatically upgraded deepseek-v4-pro to the production-ready DeepSeek-V4-Pro-0813 release, with no code changes required.

---

Details

DeepSeek's new pricing takes effect at 16:00 UTC on August 16, 2026. On the direct DeepSeek API, peak rates apply during these UTC windows:

• 01:00–04:00 UTC
• 06:00–10:00 UTC

All other hours are off-peak. Peak rates are twice the off-peak rates.

AvalAI does not increase DeepSeek V4 prices during these peak windows. We provide the off-peak rates as fixed, around-the-clock prices, so the cost of the same token type remains consistent throughout the day.

DeepSeek-V4-Pro-0813 automatic upgrade

The deepseek-v4-pro base model now automatically uses DeepSeek-V4-Pro-0813, the official production release that supersedes DeepSeek-V4-Pro Preview. Existing requests continue to use deepseek-v4-pro; no model-name or code change is required.

DeepSeek reports substantially improved agentic capabilities and stronger production performance. The release retains the V4-Pro architecture and adds a DSpark speculative-decoding module. In DeepSeek's published evaluation, V4-Pro-0813 improves over the preview on all listed benchmarks, including Terminal Bench 2.1 (87.9 vs. 72.1), DeepSWE (62.7 vs. 12.8), Toolathlon-Verified (74.1 vs. 55.9), and HLE with tools (60.0 vs. 48.2).

These figures are provider-reported benchmark results, not AvalAI guarantees. DeepSeek evaluated public code-agent tasks with the minimal DeepSeek Harness, max reasoning effort, temperature=1.0, and top_p=0.95; DSBench-FullStack and DSBench-Hard are internal DeepSeek test sets.

Fixed AvalAI pricing

Prices are in USD per 1 million tokens.

Model	Cached input	Input	Output
deepseek-v4-flash	$0.007	$0.22	$0.66
deepseek-v4-pro	$0.022	$0.66	$1.98

Comparison with direct DeepSeek peak pricing

Model	Token type	AvalAI fixed rate	Direct peak rate	Peak-hour saving with AvalAI
deepseek-v4-flash	Cached input	$0.007	$0.014	$0.007 / 1M tokens
deepseek-v4-flash	Input	$0.22	$0.44	$0.22 / 1M tokens
deepseek-v4-flash	Output	$0.66	$1.32	$0.66 / 1M tokens
deepseek-v4-pro	Cached input	$0.022	$0.044	$0.022 / 1M tokens
deepseek-v4-pro	Input	$0.66	$1.32	$0.66 / 1M tokens
deepseek-v4-pro	Output	$1.98	$3.96	$1.98 / 1M tokens

For example, 100 million deepseek-v4-pro output tokens during DeepSeek's peak hours would cost $396 through the direct DeepSeek API. The same token volume costs $198 through AvalAI, a saving of $198.

The saving depends on token type and usage volume. The comparison above is specifically against DeepSeek's direct peak-hour rates; DeepSeek's direct off-peak rates equal AvalAI's fixed rates.

Compatible model redirects

Existing integrations can continue using these identifiers. AvalAI routes them automatically to the V4 family and applies the corresponding fixed V4 price:

Existing model ID	V4 base model
deepseek-chat	deepseek-v4-flash
deepseek-coder	deepseek-v4-flash
deepseek-reasoner	deepseek-v4-pro
deepseek-v3-0324	deepseek-v4-pro
deepseek-r1-0528	deepseek-v4-pro
deepseek-v3.1	deepseek-v4-pro

No code change is required for these existing identifiers. For new integrations, use the explicit base model IDs deepseek-v4-flash and deepseek-v4-pro so the selected V4 family is clear.

API request and representative response

Request

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [
      {
        "role": "user",
        "content": "Summarize why predictable API pricing matters for high-traffic products."
      }
    ]
  }'

Representative response

{
  "id": "chatcmpl-deepseek-v4-example",
  "created": 1786876800,
  "model": "deepseek-v4-flash",
  "object": "chat.completion",
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Predictable API pricing helps high-traffic products forecast costs, set reliable budgets, and scale usage without time-of-day billing surprises."
      }
    }
  ],
  "usage": {
    "completion_tokens": 24,
    "prompt_tokens": 18,
    "total_tokens": 42,
    "prompt_tokens_details": {
      "cached_tokens": 0
    }
  },
  "estimated_cost": {
    "unit": "0.0000198000",
    "irt": 2.27,
    "exchange_rate": 114600
  }
}

The response is representative. Actual token usage, exchange rate, and estimated cost depend on the request and billing time.

What this means for AvalAI users

• Fixed DeepSeek V4 rates at every hour of the day
• Direct DeepSeek off-peak pricing even during its peak windows
• Up to 50% savings compared with direct peak-hour rates
• More predictable costs for high-traffic products
• Automatic V4 routing for existing model identifiers
• Automatic deepseek-v4-pro upgrade to DeepSeek-V4-Pro-0813 without code changes

---

Related links

• DeepSeek provider documentation
• AvalAI pricing
• Chat Completions API reference
• DeepSeek V4 model announcement