How much does the DeepSeek API cost per million tokens?

DeepSeek API pricing for V4 Flash: $0.14 input / $0.28 output per million tokens. V4 Pro: $1.74/$3.48 at standard rates, with a promotional discount periodically available (check the official page for current status). Cache hits: 1/10 of standard input on all models.

The chat interface at chat.deepseek.com is DeepSeek free to use. The API is not. It's pay-as-you-go from your topped-up balance. DeepSeek provides promotional credits to new accounts, but there is no permanent DeepSeek API free tier for production workloads. Check platform.deepseek.com for your current balance and free tier details.

How much does DeepSeek cost compared to ChatGPT and Claude?

V4 Flash is 18x cheaper than GPT-5.4 on input ($0.14 vs $2.50) and 36x cheaper than Claude Opus 4.7 ($0.14 vs $5.00). On output, Flash at $0.28 is 54x cheaper than Sonnet 4.6 at $15.00. Even V4 Pro at standard rates ($1.74) undercuts Claude Sonnet 4.6 ($3.00) on input.

What's the difference between DeepSeek V4 Flash and V4 Pro?

Flash (284B total / 13B active parameters) is the fast, cheap default for most tasks. Pro (1.6T total / 49B active) is the reasoning specialist, Codeforces 3,206, agentic coding, multi-step evaluations. Flash costs roughly 12x less than Pro at standard rates. Both support 1M context, 384K max output, and thinking/non-thinking modes.

How does DeepSeek's cache pricing work?

Automatic disk-based prefix caching. Matching prefixes serve from cache at 1/10 of standard input price. On V4 Flash, that's $0.0028 per MTok, a 98% discount. No code changes needed. The price reduction took effect April 26, 2026.

What happens to deepseek-chat and deepseek-reasoner after July 2026?

Both aliases are retired July 24, 2026, 15:59 UTC. After that, requests return errors. They currently route to V4 Flash non-thinking and thinking modes. Migration: update the model parameter to deepseek-v4-flash or deepseek-v4-pro. Same base URL, same API key. One-line change. Note: deepseek-reasoner maps to Flash, not Pro.

What is DeepSeek V4's context window?

Both V4 Flash and V4 Pro support 1M tokens with 384K max output at standard per-token rates. Previous models (V3.2, R1) maxed at 128K–164K. That's a 6–8x expansion with no per-token surcharge for long context.

How much does DeepSeek V4 cost in 2026?

DeepSeek V4 pricing has two tiers. V4 Flash costs $0.14 per million input tokens and $0.28 output — the cheapest frontier-class API available. V4 Pro costs $1.74/$3.48 at standard rates, with a 75% promotional discount periodically available that drops it to $0.435/$0.87. Both models support 1M token context and 384K max output. Cache hits cost 1/10 of standard input on both tiers.

You searched for AI Cost Optimization | Page 4 of 72

Blog

DeepSeek pricing 2026: V4, R1, API costs, and how to optimize

DeepSeek API pricing in 2026: V4 Flash at $0.14/MTok input, V4 Pro at $1.74 (promotional discount active). Cache hits at 1/10 price. Full rate table, migration deadlines, and optimization strategies.

Blog

AI ROI: How to measure and provide the return on AI investments in 2026

AI budgets are growing but proving ROI remains elusive. Learn the frameworks, metrics, and cost intelligence that integrates AI spend to business outcomes.

Blog

CloudZero AI Hub: The nexus of autonomous AI cost control

CloudZero’s AI Hub tells you why costs changed — what changed, when, who’s responsible — and how to keep them under control.

Page

Cost Per Product

Page

Cost Per Customer

Page

Cost Optimization

Blog

H100 GPU Cost In 2026: Buy, Rent, And Cloud Pricing Compared

NVIDIA H100 GPU cost in 2026: $25K–$40K new, $2–$8/hr cloud rental, used from $6K on eBay. Full buy vs. rent math, 9-provider cloud comparison, and H200/B200 context.

Blog

DevOps Best Practices: 12 Principles That Ship Faster And Cost Less In 2026

Want to streamline your software development process? Discover 7 DevOps best practices you can adopt right away.

Blog

AI observability in 2026: what it is, the five pillars, and why cost is the one everyone skips

AI observability covers performance, quality, reliability, safety, and cost. Most tools handle the first four. Here's what each pillar means, which tools cover which, and why cost is the dimension enterprises keep missing.

Blog

Best AI Governance Tools In 2026: Compliance, Security, and Cost

An honest coverage of the best AI governance tools for 2026 including CloudZero, Credo AI, OneTrust, Grip, IBM and more.