Chinese AI firm DeepSeek unveils V4 model, says it rivals U.S. giants
Key Points
- DeepSeek unveiled V4, an open‑source AI model claiming parity with Google, OpenAI and Anthropic systems.
- V4 emphasizes coding abilities, targeting developers of AI‑assisted software tools.
- The model runs on Huawei chips, highlighting China’s push for a self‑reliant AI hardware ecosystem.
- DeepSeek has not disclosed V4’s training costs or the specific hardware used.
- U.S. officials previously accused DeepSeek of using prohibited Nvidia GPUs for its R1 model.
- Anthropic alleges DeepSeek misused its Claude model to improve V4, a claim not confirmed by DeepSeek.
- Industry analysts say real‑world performance and enterprise adoption will determine V4’s impact.
DeepSeek, the Beijing‑based artificial‑intelligence startup, released a preview of its next‑generation V4 model on Friday, claiming it can match the performance of leading U.S. systems from Google, OpenAI and Anthropic. The open‑source model emphasizes coding capabilities and boasts compatibility with domestic Huawei hardware, marking a milestone for China’s chip industry. The announcement follows DeepSeek’s earlier R1 model, which drew criticism from U.S. officials over alleged use of banned Nvidia chips and accusations of misusing Anthropic’s Claude technology.
Beijing‑based DeepSeek rolled out a preview of its V4 artificial‑intelligence model on Friday, positioning the open‑source system as a direct competitor to the flagship offerings of Google, OpenAI and Anthropic. The company’s press release highlighted V4’s leap in coding proficiency, a feature that underpins popular AI agents such as ChatGPT’s Codex and Anthropic’s Claude Code.
DeepSeek’s claim rests on a series of benchmark tests that, according to the firm, place V4 on par with the closed‑source models that dominate the U.S. market. While the startup stopped short of disclosing the exact training costs or the hardware configuration used, it emphasized that V4 runs seamlessly on domestic Huawei chips, signaling a strategic push to reduce reliance on foreign semiconductor technology.
The V4 launch arrives a year after DeepSeek shocked the AI community with its R1 model, which the company said was trained for a fraction of the expense incurred by U.S. rivals. That earlier rollout sparked a wave of scrutiny from Washington. U.S. officials accused DeepSeek of employing banned Nvidia GPUs in the R1 training pipeline, a charge the firm has not addressed publicly. Anthropic, the creator of Claude, also alleged that DeepSeek leveraged its own model to accelerate development, a claim that DeepSeek has neither confirmed nor denied.
Industry observers note that the coding focus of V4 could make it attractive to developers building AI‑augmented software tools. By delivering strong performance in code generation, DeepSeek hopes to carve out a niche that complements, rather than directly replaces, existing commercial products.
Beyond the technical merits, the V4 release underscores China’s broader ambition to build a self‑sufficient AI ecosystem. The explicit mention of Huawei compatibility points to a coordinated effort between AI startups and domestic chip manufacturers, a partnership that could reshape supply‑chain dynamics in the global AI race.
DeepSeek’s announcement also touches on the evolving role of open‑source models in a market traditionally dominated by proprietary solutions. By offering V4 under an open license, the company invites external developers to experiment, adapt and potentially improve the system, a strategy that could accelerate adoption outside of China’s borders.
While the company has not released performance numbers for side‑by‑side comparisons, analysts say the real test will be how V4 handles real‑world workloads, especially in enterprise settings where reliability and security are paramount. If DeepSeek can deliver on its promises, the model may force U.S. providers to reevaluate pricing and feature strategies.
For now, the AI community watches closely as DeepSeek prepares to open V4 to a broader audience. The next few months will reveal whether the model can truly compete toe‑to‑toe with the entrenched offerings from Google, OpenAI and Anthropic, or if it will remain a regional contender bolstered by domestic chip support.