DeepSeek

DeepSeek announced the general availability of DeepSeek-V4-Pro on its official API docs, introducing major agent capability upgrades. V4-Pro now supports flexible reasoning effort controls (low for simple tasks, high for daily agent workflows, max for complex tasks) that are also extended to V4-Flash. DeepSeek added na The V4-Pro GA release also restructured DeepSeek API pricing around peak and off-peak rates, with off-peak priced at 50% of peak to enable workload scheduling flexibility. The new pricing took effect at 16:00 UTC on August 16, 2026. Existing model names were preserved, so developers did not need to migrate identifiers

Quartz reports that DeepSeek made DeepSeek-V4-Pro-0813 generally available on August 13, 2026 across app, web, and API after an April preview, with the GA version emphasizing agent capabilities such as tool use, code execution, and multi-step workflows. Reported benchmark scores include Terminal Bench 2.1 at 87.9, Deep The article documents concrete developer-facing changes: the V4-Pro API now works out of the box with the OpenAI Responses API format and includes built-in Codex integration, and thinking-effort levels for both V4-Pro and V4-Flash have been expanded to three settings (low, high, max). It also reports a scheduled V4-fam

DeepSeek has signaled a "significant" near-future price increase for its API services, according to an SCMP report dated August 6, 2026. The Hangzhou-based company posted a notice on its developer platform stating that overall pricing for API access would be raised "in the near future," with a "significant increase" ex Specific new rates have not yet been disclosed and DeepSeek says the new pricing plan will be subject to a later official notice. As of the current DeepSeek pricing page, the two available tiers are deepseek-v4-flash at $0.0028/$0.14/$0.28 per 1M tokens (cache hit input, cache miss input, output) and deepseek-v4-pro at

SiliconANGLE reports on June 16, 2026 that Chinese AI lab DeepSeek has raised more than 50 billion yuan (about $7.4 billion) in new funding, citing The Information and The Wall Street Journal, with the resulting valuation exceeding $50 billion. The figure, if confirmed, would make DeepSeek China's most valuable AI star The report frames the raise against the backdrop of intensified competition among Chinese AI providers and surging enterprise demand for low-cost frontier-tier inference, particularly for coding agents and long-context workloads where DeepSeek-V4-Pro claims open-source SOTA results. While the SiliconANGLE piece is a se

DeepSeek officially launched DeepSeek-V4 on April 24, 2026, introducing two new API model IDs: deepseek-v4-pro (1.6T total / 49B active parameters, 1M context) and deepseek-v4-flash (284B total / 13B active parameters, 1M context). Both models expose a dual Thinking / Non-Thinking mode and are accessible via the existi The V4 release note also flags a hard deprecation: the legacy aliases deepseek-chat and deepseek-reasoner will be fully retired and become inaccessible after July 24, 2026, 15:59 UTC, and until that cutoff they transparently route to deepseek-v4-flash's non-thinking and thinking modes respectively. DeepSeek states deep

Chinese startup says DeepSeek-V4-Pro beats all rival open models for maths and coding.

DeepSeek launched a developer preview of "DeepSeek Harness," an open-source agent runtime built on the Cordis plugin kernel. The framework is designed around the principle that "everything is a plugin," allowing developers to compose, swap, or extend capabilities including models, tools, skills, sessions, sandboxes, st Harness distinguishes between the model (described as "the soul of an agent") and the surrounding harness, and ships with multiple runtime modes: Standard with the full toolset, Code mode for model-orchestrated multi-round tool calls, Minimal mode that retains only a shell and file editor for clean benchmarking, and Cr