How much does 1 million tokens cost across major LLMs?

As of late 2026, input pricing per 1M tokens ranges from $0.15 on GPT-4o Mini and $0.55 on DeepSeek R1 up to $2.50 on GPT-4o and $3.00 on Claude 3.5 Sonnet. Output tokens range from $0.60/1M on mini models to $15.00/1M on Claude 3.5 Sonnet.

Why is DeepSeek R1 significantly cheaper than US frontier models?

DeepSeek R1 leverages Multi-Head Latent Attention (MLA), DeepSeekMoE architecture, and advanced FP8 mixed-precision training, dramatically reducing KV cache memory footprints and inference compute overhead.

⚡ Developer Benchmark Tool•November 2026 Official Rates

LLM Token Pricing & API Cost Estimator

Estimate your monthly AI API invoice. Adjust your anticipated prompt tokens and completion tokens to compare OpenAI, Anthropic, Google, and DeepSeek side-by-side.

Monthly Token Usage Inputs

Input Tokens (Prompts / Context):10 Million

1M25M50M100M

Output Tokens (Generated Content):5 Million

500K15M30M50M

💡 Rule of Thumb: 1 million tokens roughly equals 750,000 English words. Output tokens typically cost 3x to 5x more than input tokens across frontier models.

Projected Monthly Cost Comparison

Rates per 1M tokens

DeepSeek R1DeepSeek

$0.00/ mo

In: $0.55/1MOut: $2.19/1M

Gemini 1.5 FlashGoogle

$0.00/ mo

In: $0.07/1MOut: $0.30/1M

GPT-4o MiniOpenAI

$0.00/ mo

In: $0.15/1MOut: $0.60/1M

Gemini 1.5 ProGoogle

$0.00/ mo

In: $1.25/1MOut: $5.00/1M

Grok 2xAI

$0.00/ mo

In: $2.00/1MOut: $10.00/1M

GPT-4o (Omni)OpenAI

$0.00/ mo

In: $2.50/1MOut: $10.00/1M

Claude 3.5 SonnetAnthropic

$0.00/ mo

In: $3.00/1MOut: $15.00/1M

✓

Alex Vance

Editorial Lead•Verified: November 2026

Senior AI Software Benchmark Analyst. Over 8 years of hands-on experience stress-testing LLMs, neural image models, and developer APIs. Independent researcher with zero equity in evaluated vendors.

Read our 5-Pillar Scoring Methodology→•Direct Hands-on Tested•✉️ Contact: [email protected]

⚖️

100% Unbiased Editorial Guarantee

Strict Separation Between Commercial Relationships and Editorial Ratings

1. Zero Pay-for-Play

Software providers cannot purchase higher benchmark scores, change our verdicts, or pay to remove verified cons.

2. Self-Funded Subscriptions

We purchase independent paid tiers using our own research budget to evaluate actual paying customer experiences.

3. Monthly Pricing Audits

Our automated crawlers cross-check pricing matrices and free-tier changes every 30 days to eliminate stale info.

Spotted inaccurate pricing or altered terms?✉️ Report to Editorial Desk: [email protected]→

🔔 Price Drop Radar•Weekly Verified Deals

Never Overpay for AI Software Again

SaaS companies change their credit limits, API token pricing, and subscription tiers monthly. Get our no-fluff weekly briefing with secret discounts and benchmark shifts.

🔒 100% Privacy•Zero Spam•1-Click Unsubscribe