High-volume, cost-sensitive work
Choose Grok 4.20 Non-Reasoning
Grok 4.20 Non-Reasoning costs 21% less on a typical mix of three input tokens for every output token ($1.56 against $1.98 per million tokens).
DeepSeek V4 Pro vs Grok 4.20 Non-Reasoning: price, context and capabilities
Choose DeepSeek V4 Pro for reused prompts and documents (RAG, agents), reasoning over long documents and the latest release. Choose Grok 4.20 Non-Reasoning for high-volume, cost-sensitive work, writing and code generation and images as input.
In short
- Grok 4.20 Non-Reasoning is 21% cheaper than DeepSeek V4 Pro at the headline price ($1.56 / 1M tokens vs $1.98 / 1M tokens).
- Both accept 1M tokens per request.
- DeepSeek V4 Pro is the newer model (August 2026, against March 2026).
- Only DeepSeek V4 Pro offers reasoning.
- Only Grok 4.20 Non-Reasoning offers vision, batch api.
Which should you choose?
Each answer below comes straight from the published prices, specifications and independent benchmark results.
Writing and code generation
Choose Grok 4.20 Non-Reasoning
Grok 4.20 Non-Reasoning charges $2.50 per million output tokens, against $3.96 for DeepSeek V4 Pro. Output dominates the bill when answers are long.
Reused prompts and documents (RAG, agents)
Choose DeepSeek V4 Pro
DeepSeek V4 Pro charges $0.04 per million cached input tokens, against $0.20 for Grok 4.20 Non-Reasoning.
Reasoning over long documents
Choose DeepSeek V4 Pro
DeepSeek V4 Pro scores 80.3% on Long-context reasoning (AA-LCR), against 23.3% for Grok 4.20 Non-Reasoning.
Images as input
Choose Grok 4.20 Non-Reasoning
Grok 4.20 Non-Reasoning accepts images as input; DeepSeek V4 Pro does not.
Latest release
Choose DeepSeek V4 Pro
DeepSeek V4 Pro is the newer release (August 2026, against March 2026 for Grok 4.20 Non-Reasoning).
Monthly cost at three volumes
| Volume | DeepSeek V4 Pro | Grok 4.20 Non-Reasoning | Difference |
|---|---|---|---|
| Prototype: 1M input, 250K output tokens | $2.31 | $1.88 | Grok 4.20 Non-Reasoning, $0.44 less |
| Production: 20M input, 5M output tokens | $46.20 | $37.50 | Grok 4.20 Non-Reasoning, $8.70 less |
| Scale: 500M input, 100M output tokens | $1,056 | $875.00 | Grok 4.20 Non-Reasoning, $181.00 less |
Standard rates, without caching or batch discounts. Volumes are per month.
Specifications side by side
| DeepSeek V4 Pro | Grok 4.20 Non-Reasoning | |
|---|---|---|
| Provider | DeepSeek | xAI |
| Category | Language models | Language models |
| Status | Generally available | Generally available |
| Headline price | $1.98 / 1M tokens | $1.56 / 1M tokens |
| Input | $1.32 / 1M tokens | $1.25 / 1M tokens |
| Output | $3.96 / 1M tokens | $2.50 / 1M tokens |
| Cached input | $0.044 / 1M tokens | $0.20 / 1M tokens |
| Context window | 1M | 1M |
| Max output | 384k | No data |
| Released | August 2026 | March 2026 |
| Knowledge cutoff | No data | No data |
| Licence | Proprietary | Proprietary |
| Capabilities | Reasoning, Tool calling, Structured output, Prompt caching | Vision, Tool calling, Structured output, Prompt caching, Batch API |
| API model id | deepseek-v4-pro | grok-4.20-0309-non-reasoning |
Benchmarks side by side
Leader in bold.
| Benchmark | DeepSeek V4 Pro | Grok 4.20 Non-Reasoning |
|---|---|---|
| GPQA Diamond | 92.8% | 77.6% |
| Humanity's Last Exam | 41% | 27.9% |
| Long-context reasoning (AA-LCR) | 80.3% | 23.3% |
| Intelligence Index | 36 | No data |
| SciCode | 51% | No data |
| Terminal-Bench Hard | No data | 16.7% |
| IFBench (instruction following) | No data | 49.3% |
| τ²-Bench (agentic tool use) | No data | 59.9% |
| MMMU-Pro (visual reasoning) | No data | 64.9% |
Independent evaluations by Artificial Analysis, run the same way for every model, at the reasoning setting shown. Percentages unless marked as an index.
Frequently asked questions
Should I choose DeepSeek V4 Pro or Grok 4.20 Non-Reasoning?
Choose DeepSeek V4 Pro for reused prompts and documents (RAG, agents), reasoning over long documents and the latest release. Choose Grok 4.20 Non-Reasoning for high-volume, cost-sensitive work, writing and code generation and images as input.
Which is cheaper, DeepSeek V4 Pro or Grok 4.20 Non-Reasoning?
Grok 4.20 Non-Reasoning is 21% cheaper than DeepSeek V4 Pro at the headline price ($1.56 / 1M tokens vs $1.98 / 1M tokens).
Which has the larger context window, DeepSeek V4 Pro or Grok 4.20 Non-Reasoning?
Both accept 1M tokens per request.
What is the difference between DeepSeek V4 Pro and Grok 4.20 Non-Reasoning?
DeepSeek V4 Pro is made by DeepSeek and Grok 4.20 Non-Reasoning by xAI. Grok 4.20 Non-Reasoning is 21% cheaper than DeepSeek V4 Pro at the headline price ($1.56 / 1M tokens vs $1.98 / 1M tokens). Both accept 1M tokens per request. DeepSeek V4 Pro is the newer model (August 2026, against March 2026). Only DeepSeek V4 Pro offers reasoning. Only Grok 4.20 Non-Reasoning offers vision, batch api.
How much do DeepSeek V4 Pro and Grok 4.20 Non-Reasoning cost per month?
At 20 million input and 5 million output tokens a month, DeepSeek V4 Pro costs $46.20 and Grok 4.20 Non-Reasoning costs $37.50, at standard rates.
Which is newer, DeepSeek V4 Pro or Grok 4.20 Non-Reasoning?
DeepSeek V4 Pro was released on August 2026 and Grok 4.20 Non-Reasoning on March 2026.