DeepSeek V4 approaches: 1M tokens of context and double pricing

DeepSeek is preparing for the public release of its V4, with a 1M token window and a double pricing grid (peak/off-peak). The shadow extends beyond a simple update.

The fact

According to Tech in Asia (July 19, 2026), DeepSeek is about to finalize the launch of its V4 model. Two technical features stand out from the announcement: a context window of 1 million tokens and a two-tier pricing grid - a "peak" rate (full hours) and an "off-peak" rate (off-peak hours).

Our analysis

DeepSeek has never competed with the majors on raw capabilities alone; its leverage remains cost. Moving to 1M tokens of context is checking the box on which Anthropic (Claude Fable 5) and Moonshot (Kimi K3) build their enterprise pitch. The dual pricing goes further: it moves away from the single price per token to arbitrate on time - equivalent to Cloud Compute of EDF's peak/off-peak rates.

If the V4 delivers on these two promises at the usual DeepSeek price, it will reframe the debate on what "frontier" really costs in production - just after Kimi K3 broke the race to the lowest price.

To watch

• The official price of the two peak/off-peak rates (line-by-line comparison with K3 and Fable 5).
• Third-party benchmarks post-launch - does the 1M token window hold up on long tasks, or is it a brochure figure?