How much does DeepSeek V4 Flash cost per million tokens?

DeepSeek V4 Flash is $0.14 input / $0.28 output per million tokens as of September 2026, the cheapest published rate on any major LLM provider. V4 Flash supports a 1M-token context window with a cache hit rate of $0.0028 per million input tokens. V4 Flash is the default chat model behind DeepSeek's `deepseek-chat` API alias and the natural fit for high-volume general-purpose workloads where Sonnet 5 or GPT-5.6 Sol would otherwise dominate the bill (DeepSeek API pricing page, September 2026).

How much does DeepSeek R1 cost per million tokens?

DeepSeek R1 is $0.55 input / $2.19 output per million tokens — 73% cheaper than OpenAI o3 ($2.00/$8.00) and 10x cheaper than Claude Opus 4.6 on output ($2.19 vs $25). Cached inputs are $0.14 per million tokens, a 75% discount off standard input. The catch: R1 charges for chain-of-thought reasoning tokens at the same output rate, so a single R1 request can generate 5-15x more billed tokens than the visible answer (DeepSeek API docs, September 2026).

What is the difference between DeepSeek V3, V3.2, V4 Flash, and V4 Pro?

DeepSeek V3 at $0.27/$1.10 is the legacy general-purpose model with a 128K context window, now deprioritised in favour of V4. V3.2 at $0.28/$0.42 has the same pricing but improved coding and reasoning performance and a 32K max output. V4 Flash at $0.14/$0.28 is the new default chat model with a 1M context window. V4 Pro at $0.44/$0.87 is the flagship 1M-context model offered at 75% promotional pricing through May 31, 2026. For most production workloads, V4 Flash is the right starting point (aicostcheck.com, April 2026).

How do DeepSeek reasoning tokens work?

DeepSeek charges for chain-of-thought (CoT) reasoning tokens generated during R1 inference. These internal tokens represent the model's step-by-step thinking process, are not returned to the API caller, and are billed at the same rate as output tokens ($2.19 per million). For a typical reasoning problem with 4K input tokens, the model may generate 100 input + 150 answer output + 1,200 reasoning output = 1,450 billed tokens. The total cost is (100 × $0.55 + 1,350 × $2.19) / 1M = $0.00297 per request — 6x more expensive than V3.1 for the same task (DeployBase, March 2026).

Does DeepSeek offer prompt caching?

Yes. DeepSeek prompt caching cuts cached input costs to $0.027 per million tokens on V3 (90% off standard input) and $0.14 per million tokens on R1 (75% off standard input). Cache hits work the same way as OpenAI and Anthropic: any prompt prefix that has been seen in the cache TTL window is served at the cached rate. For workloads with stable system prompts above 1,024 tokens, the cache rate is the dominant cost lever — V4 Flash at $0.0028 cache hits is the cheapest cache rate in the LLM market (DeepSeek API docs, September 2026).

How does DeepSeek compare to OpenAI o3 and Claude Opus on cost?

DeepSeek R1 at $0.55/$2.19 is 3.7x cheaper than OpenAI o3 ($2.00/$8.00) and 10x cheaper than Claude Opus 4.6 ($5/$25) on output. DeepSeek V4 Flash at $0.14/$0.28 is 14x cheaper on input and 36x cheaper on output than GPT-5.6 Sol ($4/$20). DeepSeek V4 Pro at $0.44/$0.87 is the closest comparable to Claude Sonnet 5 ($2/$10) — 4.5x cheaper on input and 11x cheaper on output. The price gap reflects DeepSeek's open-weight model economics and aggressive Chinese-market pricing strategy (TokenMix, April 2026).

What is the DeepSeek V4 Pro promotional pricing?

DeepSeek V4 Pro launched at $0.44 input / $0.87 output per million tokens with a 75% promotional discount that ran through May 31, 2026. Standard pricing on V4 Pro is roughly $1.74 input / $3.48 output. V4 Pro supports a 1M-token context window and is DeepSeek's flagship model for enterprise workloads. After the promotional period, V4 Pro is priced competitively with Claude Sonnet 5 but remains 4-5x cheaper (LLM Cost Calculator, September 2026).

Is DeepSeek API production-ready?

Yes. DeepSeek's hosted API at api-docs.deepseek.com runs in production for major enterprise customers and supports prompt caching, batch processing (off-peak discounts), and the same OpenAI-compatible API surface used by Anthropic and OpenAI. The DeepSeek API is OpenAI-compatible: most code that targets the OpenAI Python SDK can swap the base URL and use the same model interface. The reliability trade-off is that DeepSeek is hosted in China, which raises data-residency questions for US-regulated workloads — most US customers use DeepSeek for development and cost-optimised workloads, not for sensitive production data (DeepSeek API docs, September 2026).

DeepSeek API Pricing 2026: Cost Per Million Tokens

Quick Answer

DeepSeek API pricing 2026: V4 Flash $0.14 input / $0.28 output per million tokens (cheapest major LLM); V4 Pro $0.44/$0.87 (75% promo through May 31, 2026); V3.2 $0.28/$0.42 with 128K context and 32K output; R1 $0.55/$2.19 reasoning model with separately-charged chain-of-thought tokens. Cache 90% off input.

Last verified: Sep 16, 2026.

At a glance

DeepSeek V4 Flash: $0.14 input / $0.28 output per 1M tokens (1M context; default chat)
DeepSeek V4 Pro: $0.44/$0.87 per 1M tokens (1M context; 75% promo through May 31, 2026)
DeepSeek V3.2: $0.28 input / $0.42 output per 1M tokens (128K context, 32K max output)
DeepSeek R1: $0.55/$2.19 per 1M tokens (reasoning model; CoT tokens billed as output)
Cache hit rate: $0.0028 (V4 Flash), $0.027 (V3), $0.14 (R1)
Cost vs Opus 4.6: R1 10x cheaper on output ($2.19 vs $25)
Cost vs GPT-5.6 Sol: V4 Flash 14x cheaper on input ($0.14 vs $4)

V4 Flash is the cheapest major LLM API in 2026

DeepSeek V4 Flash at $0.14 input / $0.28 output per million tokens is the cheapest published rate on any major LLM provider as of September 2026.

Compared with GPT-5.6 Sol at $4/$20, V4 Flash is 28x cheaper on input and 71x cheaper on output. Compared with Claude Sonnet 5 at $2/$10, V4 Flash is 14x cheaper on input and 36x cheaper on output. Compared with Gemini 2.5 Flash at $0.30/$2.50, V4 Flash is 2x cheaper on input and 9x cheaper on output. The price gap reflects DeepSeek's open-weight model economics, aggressive Chinese-market pricing strategy, and V4 Flash's positioning as a high-volume general-purpose workhorse (DeepSeek API pricing page, September 2026).

V4 Flash supports a 1M-token context window and runs behind DeepSeek's `deepseek-chat` API alias — the same surface that V3 ran behind before the V4 launch. Most code that targets the OpenAI Python SDK can swap the base URL to DeepSeek and use V4 Flash with no client-side changes. The reliability trade-off is that DeepSeek is hosted in China, which raises data-residency questions for US-regulated workloads (DeepSeek API docs, September 2026).

R1's reasoning premium hits output, not input

DeepSeek R1's $0.55 input / $2.19 output looks cheap, but the catch is that R1 charges for chain-of-thought reasoning tokens at the same output rate — so a single request can bill 5-15x more tokens than the visible answer.

R1 is DeepSeek's reasoning-specialist model. For a typical reasoning problem (4K input + 2K answer output), R1 may generate 1,200 reasoning tokens internally. The bill is (4,000 × $0.55 + (2,000 + 1,200) × $2.19) / 1M = $0.0075 per request. The same workload on V3.2 would cost (4,000 × $0.28 + 2,000 × $0.42) / 1M = $0.0020 per request — 3.7x more expensive on R1 despite the visible answer being only 2K tokens (DeployBase, March 2026).

The 75% cache discount on R1 inputs ($0.14 vs $0.55) is the biggest cost lever for R1 workloads. For reasoning tasks that reuse a stable system prompt across many requests, cache hits can dominate the input cost. R1 is most cost-effective when the reasoning work is genuinely hard — math, code, multi-step debugging. For routine chat or RAG, V3.2 or V4 Flash is the better economic choice (TokenMix, April 2026).

The V3, V3.2, V4 family tree

DeepSeek offers five live models in the family as of September 2026, with overlapping price-performance that can confuse first-time buyers.

V3 at $0.27 input / $1.10 output is the legacy general-purpose model with a 128K context window. The DeepSeek API now routes chat workloads to V4 Flash by default; V3 is maintained for compatibility but no longer the recommended starting point. V3.2 at $0.28 input / $0.42 output has the same input price as V3 but 2.6x cheaper output, with improved coding and reasoning performance and a 32K max output (vs V3's 16K) (DeepSeek API docs, September 2026).

V4 Flash at $0.14/$0.28 is the new default chat model. V4 Pro at $0.44/$0.87 is the flagship 1M-context model. R1 at $0.55/$2.19 is the reasoning specialist. R1-0528 at $0.50/$2.15 is a May 2025 snapshot of R1 with improved reasoning performance. For most production workloads, the recommended path is V4 Flash for general chat, V3.2 for coding, V4 Pro for enterprise flagship, and R1 for reasoning-heavy workloads (aicostcheck.com, April 2026).

Off-peak discounts and the pricing model

DeepSeek runs a peak/off-peak pricing model that cuts the bill by 50% during off-peak hours.

DeepSeek's pricing model charges 50% less during off-peak hours for the same workload. Off-peak hours are typically 16:30-00:30 UTC (00:30-08:30 Beijing time). The model is opt-in: developers can choose to route traffic during off-peak hours by setting the appropriate parameters. For batch processing, embeddings generation, and other non-real-time workloads, off-peak routing cuts the bill by 50% on top of any prompt-cache savings (DeepSeek API docs, September 2026).

The 50% off-peak discount is the same magnitude as OpenAI Batch API (50% off standard) and Google Vertex AI Batch Mode (50% off standard). DeepSeek's pricing model does not include a separate Batch API — off-peak routing is the equivalent mechanism. For DeepSeek V4 Flash, the off-peak rate is $0.07 input / $0.14 output per million tokens — cheaper than any other provider's standard rate (DeepSeek API docs, September 2026).

What enterprise buyers should do next

Three actions for organisations evaluating DeepSeek API pricing in 2026.

Pick V4 Flash as the default. For most production workloads, V4 Flash is the right starting point. Move to V3.2 when the task needs better coding or longer output (32K). Move to R1 when the task needs deeper reasoning and the cost of CoT tokens is justified by accuracy gains. Move to V4 Pro when you need 1M context and enterprise SLAs.
Use prompt caching aggressively. DeepSeek's cache rate of $0.0028 per million tokens on V4 Flash is the cheapest in the LLM market. A 50K-token system prompt reused 100x per hour costs $0.014 per hour on V4 Flash versus $0.50 per hour on V3 without caching.
Plan for the China data-residency question. DeepSeek is hosted in China. For US-regulated workloads (HIPAA, FedRAMP, CMMC), use DeepSeek for development and cost-optimised non-sensitive workloads, but route sensitive production data through Azure, Bedrock, or Anthropic direct. For non-regulated workloads, DeepSeek's 14-71x cost advantage is real and durable.

What to watch next

Three near-term datapoints. First, DeepSeek's announced V4 Pro API service after May 31, 2026 — the 75% promotional discount ended May 31, 2026, and standard pricing applies since then. Second, the introduction of DeepSeek on AWS Bedrock and Azure AI Foundry as a hosted model — currently DeepSeek is direct-only. Third, the release of DeepSeek R2 — DeepSeek typically ships a successor reasoning model every 8-12 months; R2 is plausible for Q4 2026.

#deepseek api pricing 2026 #deepseek v4 flash #deepseek v3.2 #deepseek r1 pricing #reasoning tokens cost #cheap llm api

Written by

Fazlur RahmanAI, Entrepreneurship, EdTech

Fazlur Rahman is the founder of Tutorsbot, building AI-powered tools for learning and career growth. He writes about applying AI in real products and the practi… Read more

Fazlur Rahman is the founder of Tutorsbot, building AI-powered tools for learning and career growth. He writes about applying AI in real products and the practical side of building an ed-tech startup.