DeepSeek-V4-Pro-0813 (general availability)
V4-Pro goes GA with large agentic gains (Terminal Bench 2.1: 87.9, DeepSWE 62.7), low/high/max effort, Responses API and 50% off-peak pricing.
Agent-focused post-training of V4-Pro, DSpark built in, native Responses API for Codex. App adds 'Expert Mode'. From 2026-08-16 the API uses peak/off-peak pricing, with off-peak at half the peak rate.
- Date
- Thursday, 13 August 2026
- Lab
- DeepSeek
- Kind
- open-weights
- Access
- open weights
- Price
- From 2026-08-16: Pro cache-hit input $0.044 peak / $0.022 off-peak per M; output $3.96 peak / $1.98 off-peak (pricing page, fetched 2026-10-04)
Figures
| Measure | Value | Measured by |
|---|---|---|
| Terminal Bench 2.1 | 87.9 Toolathlon-Verified 74.1; DeepSWE 62.7; DSBench-FullStack 71.1 | company |
Peak hours are 01:00-04:00 and 06:00-10:00 UTC Mon-Fri excluding Chinese holidays. HF card lists 1.7T parameters vs 1.6T for the preview. Pricing page also notes V4-Pro does not support vision.
Sources
- api-docs.deepseek.com/news/news260813
- huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813
- api-docs.deepseek.com/quick_start/pricing
This record was checked against its sources on 6 October 2026. How we check