DeepSeek V4 vs Grok 4.3
Benchmarks, API pricing and specs, head to head. Data updated 2026-06-10.
DeepSeek · Mar 2026
The open-weights shock of 2026: ~81% SWE-bench Verified, a 1M context and MIT license at $0.30/$0.50 per million tokens.
xAI · Apr 2026
xAI's frontier model: 1M context at $1.25/$2.50, ~168 tokens/sec, and a record low hallucination rate on independent testing.
The verdict
These two models don't yet share verified results on the benchmarks we track, so judge them on specs, pricing and intended use. On price the gap is dramatic: DeepSeek V4 works out roughly 4.5x cheaper per blended million tokens. And DeepSeek V4 is open-weights (MIT), so it can be self-hosted — a structural advantage if data control or vendor independence matters.
Specs & pricing
| DeepSeek V4 | Grok 4.3 | |
|---|---|---|
| modhub Index | — | — |
| Input price / 1M | $0.3 | $1.25 |
| Output price / 1M | $0.5 | $2.5 |
| Context window | 1M | 1M |
| Max output | 64K | — |
| Open weights | yes (MIT) | no |
| Reasoning model | yes | yes |
| Multimodal input | text | text, image |
| Knowledge cutoff | Dec 2025 | Mar 2026 |
| Released | Mar 2026 | Apr 2026 |
| Example monthly cost* | $3.75 | $16.25 |
* 10M input + 1.5M output tokens per month at list prices, no caching. Green = better value on that row.
Frequently asked questions
- Which is better, DeepSeek V4 or Grok 4.3?
- These two models don't yet share verified results on the benchmarks we track, so judge them on specs, pricing and intended use. On price the gap is dramatic: DeepSeek V4 works out roughly 4.5x cheaper per blended million tokens. And DeepSeek V4 is open-weights (MIT), so it can be self-hosted — a structural advantage if data control or vendor independence matters.
- Which is cheaper, DeepSeek V4 or Grok 4.3?
- DeepSeek V4 costs $0.3/$0.5 per million input/output tokens, while Grok 4.3 costs $1.25/$2.5. For a typical workload of 10M input and 1.5M output tokens per month, that's $3.75 versus $16.25.
- Which model is better for coding, DeepSeek V4 or Grok 4.3?
- We don't yet track SWE-bench Verified results for both models; check their individual pages for coding-related scores.