DeepSeek Launches V4 Pro API With 1M-Token Context Window and Agent-First Tooling

DeepSeek has pushed its V4 Pro model to general availability, listing the release as DeepSeek-V4-Pro-0813 on its pricing page. The flagship tier ships with a 1M-token context window and a 384K-token maximum output, defaulting to a thinking mode while still exposing a non-thinking endpoint for latency-sensitive calls.

For developers building on long codebases, large document corpora, or multi-step agent runs, the combination of 1M input and 384K output means a single API call can ingest and generate far more material than the previous generation permitted. DeepSeek has also built in the capabilities most agent developers now expect: structured JSON output, tool calls, the Responses API, and Anthropic-API compatibility, plus beta support for conversation-prefix continuation and FIM completion — with FIM restricted to the non-thinking mode. At the wire level, the service exposes both OpenAI- and Anthropic-compatible endpoints, meaning teams already running on either standard can switch without re-plumbing their stack.

This points to a meaningful shift in how DeepSeek is positioning its model family. Rather than offering a single general-purpose tier, the V4 lineup now divides cleanly between a high-throughput Flash layer and a heavier-reasoning Pro layer — a structure that mirrors how larger frontier labs have organised their own model families. The practical effect for enterprise buyers is a clearer decision matrix: route cost-sensitive, high-frequency traffic to Flash and reserve Pro for the workloads that genuinely require longer context and richer agent capabilities.

Pricing reflects that deliberate split. Pro is set at 0.025 yuan per million tokens for cache hits, 3 yuan for cache misses on input, and 6 yuan on output. Flash, by contrast, is priced at 0.02, 1, and 2 yuan respectively — putting Pro at roughly three times the cost of Flash on cache-miss input and output. Concurrency limits reinforce the divide: Flash supports 2,500 concurrent requests, while Pro is capped at 500.

Even at the Pro tier's elevated rate, the absolute price remains modest by frontier-model standards. This follows a broader pattern of DeepSeek using aggressive pricing to undercut most overseas competitors while still introducing tiered structure that gives the company room to recoup costs following a recent price-increase announcement. The two-track V4 lineup — low-cost, high-concurrency Flash for scale, and higher-spec Pro aimed at coding and agent use cases — is now effectively complete and live for API consumers.