DeepSeek V4: The Open-Source Model That Frontier Labs Feared
agents claude deepseek open-source
| Source: HN | Original article
DeepSeek V4, an open-source model, outperforms rivals at a fraction of the cost.
DeepSeek V4 has been released as an open-source model, marking a significant shift in the AI landscape. This new model ships under the MIT license with a significantly lower cost of $0.30 per million output tokens, making it 83 times cheaper than Claude Opus 4.7. Despite its lower cost, DeepSeek V4 achieves impressive performance, scoring 80.6% on the SWE-bench Verified benchmark.
This development matters because it challenges the dominance of closed-source frontier labs, which have long been the standard for high-performance AI models. The release of DeepSeek V4 threatens to disrupt the industry by providing near-frontier performance at a fraction of the cost of traditional models. As we reported on May 13, the company behind the GLiNER model had already released an open-source model for running LLM guardrails, but DeepSeek V4 takes this a step further by offering a trillion-parameter model that can run on non-Nvidia hardware.
As the AI landscape continues to evolve, it will be interesting to watch how DeepSeek V4 compares to other frontier models, such as GPT-5.5 and Grok 5, which are expected to ship in Q2 2026. With its open-source nature and lower cost, DeepSeek V4 has the potential to democratize access to high-performance AI models, and its impact on the industry will be closely watched in the coming months.
Sources
Back to AIPULSEN