Together AI Adds DeepSeek V4 Pro to Cloud Platform, Targeting Long-Context Coding Workloads

According to a recent LinkedIn post from Together AI, the company is highlighting the availability of DeepSeek V4 Pro on its AI Native Cloud platform. The post describes the model as a long-context reasoning and coding system aimed at AI-native users, with support for agentic workflows and long-horizon coding tasks.

The company’s LinkedIn post highlights several technical benchmarks, including 93.5% on LiveCodeBench, a Codeforces rating of 3206, and 80.6% on SWE-Bench Verified, suggesting competitive coding performance. It also cites efficiency gains over a prior V3.2 version, including reduced FLOPs and KV cache usage for long-context inference.

As shared in the post, DeepSeek V4 Pro offers three reasoning modes—Non-think, Think High, and Think Max—allowing users to trade off speed and depth of reasoning. The model is presented as production-ready with a 99.9% SLA on serverless deployment and support for a 512K token context window, with a related DeepSeek V4 Flash model described as coming soon.

For investors, the post suggests that Together AI is strengthening its position as an infrastructure provider for advanced code-focused and long-context AI applications. If adoption of DeepSeek V4 Pro on Together’s platform scales, this could support higher usage-based revenue, deepen relationships with AI-native developers, and enhance the company’s competitiveness against other model-hosting and inference providers.

The emphasis on reliability SLAs, efficiency improvements, and differentiated reasoning modes may appeal to enterprise and developer customers seeking cost-effective, production-grade AI capabilities. In a crowded AI infrastructure market, such additions to the model catalog could help Together AI capture workloads that require high-context reasoning and coding performance, potentially improving its longer-term growth profile.