High-volume, cost-sensitive work
Choose DeepSeek V4 Pro
DeepSeek V4 Pro costs 1% less on a typical mix of three input tokens for every output token ($1.98 against $2.00 per million tokens).
Choose Claude Haiku 4.5 for images as input. Choose DeepSeek V4 Pro for high-volume, cost-sensitive work, writing and code generation, reused prompts and documents (RAG, agents), very long inputs and the latest release.
Each answer below comes straight from the published prices, specifications and independent benchmark results.
Choose DeepSeek V4 Pro
DeepSeek V4 Pro costs 1% less on a typical mix of three input tokens for every output token ($1.98 against $2.00 per million tokens).
Choose DeepSeek V4 Pro
DeepSeek V4 Pro charges $3.96 per million output tokens, against $5.00 for Claude Haiku 4.5. Output dominates the bill when answers are long.
Choose DeepSeek V4 Pro
DeepSeek V4 Pro charges $0.04 per million cached input tokens, against $0.10 for Claude Haiku 4.5.
Choose DeepSeek V4 Pro
DeepSeek V4 Pro reads up to 1M tokens per request, against 200k for Claude Haiku 4.5.
Choose Claude Haiku 4.5
Claude Haiku 4.5 accepts images as input; DeepSeek V4 Pro does not.
Choose DeepSeek V4 Pro
DeepSeek V4 Pro is the newer release (August 2026, against October 2025 for Claude Haiku 4.5).
| Volume | Claude Haiku 4.5 | DeepSeek V4 Pro | Difference |
|---|---|---|---|
| Prototype: 1M input, 250K output tokens | $2.25 | $2.31 | Claude Haiku 4.5, $0.06 less |
| Production: 20M input, 5M output tokens | $45.00 | $46.20 | Claude Haiku 4.5, $1.20 less |
| Scale: 500M input, 100M output tokens | $1,000 | $1,056 | Claude Haiku 4.5, $56.00 less |
Standard rates, without caching or batch discounts. Volumes are per month.
| Claude Haiku 4.5 | DeepSeek V4 Pro | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Category | Language models | Language models |
| Status | Legacy | Generally available |
| Headline price | $2.00 / 1M tokens | $1.98 / 1M tokens |
| Input | $1.00 / 1M tokens | $1.32 / 1M tokens |
| Output | $5.00 / 1M tokens | $3.96 / 1M tokens |
| Cached input | $0.10 / 1M tokens | $0.044 / 1M tokens |
| Context window | 200k | 1M |
| Max output | 64k | 384k |
| Released | October 2025 | August 2026 |
| Knowledge cutoff | No data | No data |
| Licence | Proprietary | Proprietary |
| Capabilities | Vision, Reasoning, Tool calling, Structured output, Prompt caching, Batch API | Reasoning, Tool calling, Structured output, Prompt caching |
| API model id | claude-haiku-4-5-20251001 | deepseek-v4-pro |
Leader in bold.
| Benchmark | Claude Haiku 4.5 | DeepSeek V4 Pro |
|---|---|---|
| Intelligence Index | No data | 36 |
| SciCode | No data | 51% |
| GPQA Diamond | No data | 92.8% |
| Humanity's Last Exam | No data | 41% |
| Long-context reasoning (AA-LCR) | No data | 80.3% |
Independent evaluations by Artificial Analysis, run the same way for every model, at the reasoning setting shown. Percentages unless marked as an index.
Choose Claude Haiku 4.5 for images as input. Choose DeepSeek V4 Pro for high-volume, cost-sensitive work, writing and code generation, reused prompts and documents (RAG, agents), very long inputs and the latest release.
DeepSeek V4 Pro is 1% cheaper than Claude Haiku 4.5 at the headline price ($1.98 / 1M tokens vs $2.00 / 1M tokens).
DeepSeek V4 Pro reads more per request: 1M tokens against 200k.
Claude Haiku 4.5 is made by Anthropic and DeepSeek V4 Pro by DeepSeek. DeepSeek V4 Pro is 1% cheaper than Claude Haiku 4.5 at the headline price ($1.98 / 1M tokens vs $2.00 / 1M tokens). DeepSeek V4 Pro reads more per request: 1M tokens against 200k. DeepSeek V4 Pro is the newer model (August 2026, against October 2025). Only Claude Haiku 4.5 offers vision, batch api.
At 20 million input and 5 million output tokens a month, Claude Haiku 4.5 costs $45.00 and DeepSeek V4 Pro costs $46.20, at standard rates.
DeepSeek V4 Pro was released on August 2026 and Claude Haiku 4.5 on October 2025.