Grok 4.7 and DeepSeek V4 Pro both sit in the lower-cost tier of capable models. Grok 4.7 costs $2 per million input tokens and $6 per million output tokens, with a 500,000-token context window and image input. DeepSeek V4 Pro costs $1.32 and $3.96 at peak on DeepSeek's API, and half that off-peak, with a 1M-token window and text-only input.
They differ on where the model runs and what it can see. V4 Pro's weights are on Hugging Face under the MIT licence, so you can host it yourself. Grok 4.7 is only available from SpaceXAI (the name xAI now uses) and partners such as Cursor. Grok has the live-data angle: the Grok app answers with information from the web and X, and the API offers Web Search and X Search tools. SpaceXAI aims Grok 4.7 at long-running coding and office documents, and DeepSeek says V4-Pro reaches top-tier results on coding benchmarks. Test both on your own tasks in Omny.
Which should you choose?
Grok
Grok 4.7 · 500K context
Pick it if
you want X Search, image input or Grok inside Cursor and Grok Build.
DeepSeek
DeepSeek V4 Pro 0423 · 1M context
Pick it if
you need MIT-licensed weights or the lower per-token cost.
Pick Grok 4.7 for image input, X Search and ready integrations in Cursor and Grok Build. Pick DeepSeek V4 Pro for open weights, self-hosting and the lower token price.
Questions
Can DeepSeek access real-time information?
Not by itself. DeepSeek's API supports tool calls but lists no built-in web search, so live data has to come from tools you provide. Grok's app answers with information from the web and X, and SpaceXAI's API offers Web Search and X Search tools.
Which is cheaper to run at scale?
At list prices DeepSeek V4 Pro is cheaper than Grok 4.7, and its off-peak rates are half its peak rates. With open weights you can also run it on your own hardware, though a 1.6T-parameter model needs a large deployment.