You are trying to ship a feature that burns through tokens fast, and the invoice from last month made you flinch. The choice between xAI and DeepSeek is not just about capability; it is about whether your architecture can survive a 15x difference in input costs. DeepSeek’s flash models start at $0.065 per 1M input tokens, while xAI’s entry-level grok-build-0.1 charges $1 for the same volume. That is not a rounding error; it is a different financial reality for anyone running high-volume inference.
### The gap in the ledger
| Vendor | Model | Input | Cached input | Output |
|--------|-------|-------|--------------|--------|
| xAI | grok‑build‑0.1 | $1 per 1M tokens | $0.2 per 1M tokens | $2 per 1M tokens |
| xAI | grok‑4.3 | $1.25 per 1M tokens | $0.2 per 1M tokens | $2.5 per 1M tokens |
| xAI | grok‑4.20‑multi‑agent | $1.25 per 1M tokens | $0.2 per 1M tokens | $2.5 per 1M tokens |
| xAI | grok‑4.20 | $1.25 per 1M tokens | $0.2 per 1M tokens | $2.5 per 1M tokens |
| xAI | grok‑4.6 | $2 per 1M tokens | $0.5 per 1M tokens | $6 per 1M tokens |
| xAI | grok‑4.5 | $2 per 1M tokens | $0.3 per 1M tokens | $6 per 1M tokens |
| DeepSeek | deepseek‑v4‑flash‑0731 | $0.065 per 1M tokens | $0.016 per 1M tokens | $0.18 per 1M tokens |
| DeepSeek | deepseek‑v4‑flash | $0.0668 per 1M tokens | $0.0134 per 1M tokens | $0.1336 per 1M tokens |
| DeepSeek | deepseek‑v4.1‑flash | $0.15 per 1M tokens | $0.003 per 1M tokens | $0.6 per 1M tokens |
| DeepSeek | deepseek‑v4‑flash‑vision‑exp | $0.22 per 1M tokens | $0.007 per 1M tokens | $0.66 per 1M tokens |
| DeepSeek | deepseek‑chat‑v3.1 | $0.25 per 1M tokens | $0.13 per 1M tokens | $0.95 per 1M tokens |
| DeepSeek | deepseek‑chat‑v3‑0324 | $0.25 per 1M tokens | – | $1 per 1M tokens |
| DeepSeek | deepseek‑chat | $0.2574 per 1M tokens | – | $1.0287 per 1M tokens |
| DeepSeek | deepseek‑v3.2 | $0.269 per 1M tokens | $0.1345 per 1M tokens | $0.4 per 1M tokens |
| DeepSeek | deepseek‑v3.2‑exp | $0.27 per 1M tokens | – | $0.41 per 1M tokens |
| DeepSeek | deepseek‑v3.1‑terminus | $0.27 per 1M tokens | $0.135 per 1M tokens | $1 per 1M tokens |
The table reveals a stark divide. DeepSeek’s flash series, including deepseek-v4-flash-0731 and deepseek-v4-flash, keeps output costs under $0.18 per 1M tokens, a fraction of what xAI charges for its mid-tier models. Even DeepSeek’s more expensive chat models, like deepseek-chat-v3.1 at $0.25 input and $0.95 output, remain significantly cheaper than xAI’s grok-4.5 or grok-4.6, which charge $2 for input and $6 for output. For teams building chatbots or RAG pipelines where prompt caching is a major strategy, DeepSeek’s cached input rates, such as $0.003 for deepseek-v4.1-flash, offer a level of cost efficiency that xAI’s flat $0.2 cached rate cannot match. xAI’s pricing structure suggests a premium positioning, likely tied to specific performance benchmarks or orchestration features, but the raw cost per token is undeniably higher across the board.
If your primary constraint is budget and you do not require specific xAI proprietary tools, DeepSeek is the clear winner. It allows you to scale usage without the exponential cost growth seen in xAI’s lineup. However, if your application relies heavily on the specific multi-agent orchestration capabilities of grok-4.20-multi-agent or the higher output quality of grok-4.6, you must accept the premium. For most developers, though, the economic argument is lopsided. Pick DeepSeek for cost-sensitive, high-volume workloads. Choose xAI only if the specific model capabilities justify the 15x to 30x price difference. Pricing verified on 2026-09-12.