DeepSeek runs two API models, V4-Pro and V4.1-Flash, both with a 1M context window. Its prices depend on the time of day. The rates in our tables are off-peak, and weekday peak hours cost twice as much.
DeepSeek V4.1-Flash costs $0.15 input / $0.60 output per 1M tokens off-peak, and V4-Pro $0.66 / $1.98. Peak hours (01:00–04:00 and 06:00–10:00 UTC, Monday to Friday) double every rate. Cache hits cost a small fraction of the input price: $0.003 on Flash.
| Model | Tier | Input / 1M | Output / 1M | Cache hit / 1M | Context | Notes |
|---|---|---|---|---|---|---|
| DeepSeek V4-Pro | Mid-tier | $0.66 | $1.98 | $0.022 | 1M | A planned Sep 14, 2026 reroute to V4.1-Flash rates was reversed after user pushback; billing unchanged since the Aug 13, 2026 GA build. Off-peak rate shown; peak (01:00-04:00 & 06:00-10:00 UTC weekdays) is 2x: $1.32/$3.96 |
| DeepSeek V4.1-Flash | Budget | $0.15 | $0.60 | $0.003 | 1M | Released Sep 10, 2026, replacing V4-Flash & V4-Flash-Vision-Exp at a lower price; off-peak rate shown. Peak (01:00-04:00 & 06:00-10:00 UTC weekdays) is 2x: $0.30/$1.20 |
| Model | Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens |
|---|---|---|---|
| DeepSeek V4.1-Flash | $5.40 | $4.98 | $0.16 |
| DeepSeek V4-Pro | $21.12 | $21.38 | $0.64 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
| Model | Provider | Tier | Input / 1M | Output / 1M | Context |
|---|---|---|---|---|---|
| DeepSeek V4-Pro | DeepSeek | Mid-tier | $0.66 | $1.98 | 1M |
| Mistral Large 3 | Mistral | Mid-tier | $0.50 | $1.50 | 256K |
| Mistral Large 4 | Mistral | Mid-tier | $0.70 | $2.10 | 1M |
| Grok 4.3 | SpaceXAI | Mid-tier | $1.25 | $2.50 | 1M |
| Gemini 3.8 Flash | Mid-tier | $0.75 | $3.75 | 1M | |
| DeepSeek V4.1-Flash | DeepSeek | Budget | $0.15 | $0.60 | 1M |
| Ministral 3 14B | Mistral | Budget | $0.20 | $0.20 | 256K |
| Gemini 2.5 Flash-Lite | Budget | $0.10 | $0.40 | 1M | |
| GPT-6 Luna | OpenAI | Budget | $0.10 | $0.50 | 1M |
| Claude Haiku 5.5 | Anthropic | Budget | $0.10 | $0.50 | 1M |
DeepSeek's cheapest model per tier (bold) against the four cheapest same-tier models from other providers, by output price.
DeepSeek's off-peak rates are exactly half its peak rates. Peak runs 01:00–04:00 and 06:00–10:00 UTC on weekdays, and weekends and Chinese public holidays are off-peak all day. In practice that's the Asian working day and early European morning. Interactive traffic from US users mostly lands off-peak already. Batch-style jobs such as data backfills, evals and bulk summaries should be queued to start after 10:00 UTC, or on weekends, to keep them at the rates shown here.
A cache hit on V4.1-Flash costs $0.003 per 1M against $0.15 for a miss, and on V4-Pro $0.022 against $0.66. For agents and chat apps that resend the same system prompt and history, the share of input that hits the cache matters more than the headline rate. Keep the stable parts of the prompt identical and at the front. Avoid putting timestamps or request ids at the top of the system prompt.
Call V4.1-Flash as deepseek-flash. The legacy ids deepseek-v4-flash and deepseek-v4-flash-vision-exp still work, but they're served by V4.1-Flash and billed at its price. Both models support thinking and non-thinking modes, JSON output, tool calls and an Anthropic-compatible API format. Only Flash accepts images. DeepSeek lists a concurrency limit of 2,500 for Flash and 500 for V4-Pro, which matters more than price if you plan to fan out many parallel agent calls on Pro.
DeepSeek's built-in Search and Fetch tools are charged per successful request on top of tokens. Calls made through DeepSeek Harness, Claude Code or Codex were free until per-request billing for them began on Nov 1, 2026 (00:00 Beijing time). As with token charges, DeepSeek deducts from your granted balance before your topped-up balance. If your agent searches on most turns, count those calls in your cost model.
In the mid tier, clearly: V4-Pro at $0.66 / $1.98 off-peak costs far less than GPT-6.1 Sol or Claude Sonnet 5.5 ($2.00 / $10.00). At the budget end it isn't. GPT-6 Luna and Claude Haiku 5.5 both list $0.10 / $0.50, slightly under V4.1-Flash's off-peak rate and well under its peak rate. Flash still wins on cache hits.
Peak is 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday, excluding Chinese public holidays. Every other hour is off-peak, and off-peak rates are half the peak rates. The off-peak V4-Pro rate is $1.98 per 1M output tokens; at peak it's $3.96.
It was replaced by DeepSeek V4.1-Flash on Sep 10, 2026. The old id still works, but it's served by V4.1-Flash and billed at the V4.1-Flash price, so existing code got cheaper without any changes.
Yes. The DeepSeek web and mobile chat app has no paid plan in our data. The API is billed separately, per token, from a prepaid balance.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.