Reference Data

AI API Pricing Index — July 2026

Updated pricing across every major AI provider. Input tokens, output tokens, context windows, and special features. Downloadable dataset. Last updated July 11, 2026.

Start Free → Plugsky pricing Cost calculator

Methodology

Prices listed in this index were collected from each provider's official pricing page on July 11, 2026. All prices are in USD per million tokens (MTok) for API access. Batch and volume discount pricing is not included. Where providers offer different rates for different context lengths (e.g., cached vs. uncached), the standard real-time rate is shown.

Official pricing sources:

Full pricing table

31 models across 11 providers. Prices in USD per million tokens (MTok). Context in tokens. Plugsky models are flat-rate unlimited — highlighted below.

Provider Model Input ($/MTok) Output ($/MTok) Context Data cutoff Notes
OpenAIGPT-4o2.5010.00128KOct 2025Vision, structured outputs
OpenAIGPT-4o-mini0.150.60128KOct 2025Fast, cheap, multimodal
OpenAIGPT-430.0060.008KSep 2021Legacy — deprecated
OpenAIo115.0060.00128KOct 2025Reasoning, chain-of-thought
OpenAIo3-mini1.104.40128KOct 2025Fast reasoning
AnthropicClaude 3.5 Sonnet3.0015.00200KDec 2025Top-tier, vision, code
AnthropicClaude 3.5 Haiku0.804.00200KDec 2025Fast, cheap
AnthropicClaude 3 Opus15.0075.00200KAug 2024Max intelligence
GoogleGemini 2.5 Pro1.255.001MMay 20261M context, code, audio
GoogleGemini 2.5 Flash0.150.601MMay 2026Fast, multimodal, 1M ctx
GoogleGemini 2.0 Flash0.100.401MFeb 2025Cheapest per-token
xAIGrok-35.0015.00128KMar 2026Reasoning, real-time
xAIGrok-22.0010.00128KAug 2024Legacy model
DeepSeekDeepSeek-V40.502.00128KJun 2026MoE, strong reasoning
DeepSeekDeepSeek-Coder-V40.281.10128KJun 2026Code-specialised
MistralMistral Large 34.0012.00128KJan 2026Multilingual, reasoning
MistralMistral Small 40.200.6032KJan 2026Fast, lightweight
MistralMistral Medium 3.52.708.10128KJan 2026Mid-range balance
GroqLlama-3.3-70B0.590.79128KDec 2025LPU inference, fast
GroqMixtral-8x7B0.240.2432KDec 2024Legacy MoE
Together AILlama-3.3-70B0.880.88128KDec 2025Hosted open model
Together AIMixtral-8x22B1.201.2064KDec 2024MoE hosted
Fireworks AILlama-3.3-70B0.700.70128KDec 2025Serverless hosted
Fireworks AIMixtral-8x7B0.500.5032KDec 2024Serverless hosted
CerebrasLlama-3.3-70B0.600.60128KDec 2025Wafer-scale, fast
Plugskyplugsky-microFree (trial)31+—Flat rate
Plugskyplugsky-lite$20/mo flat31+—Flat rate
Plugskyplugsky-plus$60/mo flat31+—Flat rate
Plugskyplugsky-pro$60/mo flat31+—Flat rate
Plugskyplugsky-max$60/mo flat31+—Flat rate
Plugskyplugsky-frontier$120/mo flat31+—Flat rate

Plugsky models shown with flat-rate plans — unlimited tokens, no per-token metering. "31+" refers to the full model catalog available per plan. Prices as of July 11, 2026.

Download

The data above is also available in CSV format. Copy the block below and paste into any spreadsheet application:

Provider,Model,Input ($/MTok),Output ($/MTok),Context,Data cutoff,Notes
OpenAI,GPT-4o,2.50,10.00,128K,Oct 2025,Vision, structured outputs
OpenAI,GPT-4o-mini,0.15,0.60,128K,Oct 2025,Fast, cheap, multimodal
OpenAI,GPT-4,30.00,60.00,8K,Sep 2021,Legacy - deprecated
OpenAI,o1,15.00,60.00,128K,Oct 2025,Reasoning chain-of-thought
OpenAI,o3-mini,1.10,4.40,128K,Oct 2025,Fast reasoning
Anthropic,Claude 3.5 Sonnet,3.00,15.00,200K,Dec 2025,Top-tier vision code
Anthropic,Claude 3.5 Haiku,0.80,4.00,200K,Dec 2025,Fast cheap
Anthropic,Claude 3 Opus,15.00,75.00,200K,Aug 2024,Max intelligence
Google,Gemini 2.5 Pro,1.25,5.00,1M,May 2026,1M context code audio
Google,Gemini 2.5 Flash,0.15,0.60,1M,May 2026,Fast multimodal 1M ctx
Google,Gemini 2.0 Flash,0.10,0.40,1M,Feb 2025,Cheapest per-token
xAI,Grok-3,5.00,15.00,128K,Mar 2026,Reasoning real-time
xAI,Grok-2,2.00,10.00,128K,Aug 2024,Legacy model
DeepSeek,DeepSeek-V4,0.50,2.00,128K,Jun 2026,MoE strong reasoning
DeepSeek,DeepSeek-Coder-V4,0.28,1.10,128K,Jun 2026,Code-specialised
Mistral,Mistral Large 3,4.00,12.00,128K,Jan 2026,Multilingual reasoning
Mistral,Mistral Small 4,0.20,0.60,32K,Jan 2026,Fast lightweight
Mistral,Mistral Medium 3.5,2.70,8.10,128K,Jan 2026,Mid-range balance
Groq,Llama-3.3-70B,0.59,0.79,128K,Dec 2025,LPU inference fast
Groq,Mixtral-8x7B,0.24,0.24,32K,Dec 2024,Legacy MoE
Together AI,Llama-3.3-70B,0.88,0.88,128K,Dec 2025,Hosted open model
Together AI,Mixtral-8x22B,1.20,1.20,64K,Dec 2024,MoE hosted
Fireworks AI,Llama-3.3-70B,0.70,0.70,128K,Dec 2025,Serverless hosted
Fireworks AI,Mixtral-8x7B,0.50,0.50,32K,Dec 2024,Serverless hosted
Cerebras,Llama-3.3-70B,0.60,0.60,128K,Dec 2025,Wafer-scale fast
Plugsky,plugsky-micro,Free (trial),,,,Unlimited usage
Plugsky,plugsky-lite,$20/mo flat,,,,,Unlimited usage
Plugsky,plugsky-plus,$60/mo flat,,,,,Unlimited usage
Plugsky,plugsky-pro,$60/mo flat,,,,,Unlimited usage
Plugsky,plugsky-max,$60/mo flat,,,,,Unlimited usage
Plugsky,plugsky-frontier,$120/mo flat,,,,,Unlimited usage

Last updated: July 11, 2026. Check individual provider pages for the latest rates.

AI API pricing has been on a steep downward trajectory. Since mid-2024, prices-per-token have dropped approximately 50% year-over-year across most providers. Key trends:

  • Commoditisation of small models: Sub-$0.20/MTok inference is now available from Google (2.0 Flash), DeepSeek (Coder-V4), and Mistral (Small 4).
  • Flat-rate disruption: Plugsky's unlimited-token model is pushing per-token pricing downward. Expect more providers to experiment with flat monthly tiers.
  • Frontier models hold value: GPT-4o, Claude 3.5 Sonnet, and Mistral Large 3 remain above $2/MTok input — the premium for top-tier intelligence is stable.
  • Context inflation: 1M-token context windows (Google Gemini) are becoming standard on new models, up from 128K-200K in 2025.
  • Reasoning premium: Models with chain-of-thought (o1, Grok-3) carry 3-6x output premiums over standard models.

Expect continued price compression in 2027 as open-weight models improve and inference hardware becomes more efficient.

How to use this data

  • Budget planning: Estimate monthly token consumption (input + output) × per-token price = baseline cost. Add 20-30% for variance.
  • Provider comparison: Evaluate total cost of ownership including latency, availability, data residency, and compliance — not just per-token price.
  • Model selection: Match model capability to task complexity. Use cheap models for classification, extraction, and routing; frontier models for reasoning, generation, and analysis.
  • Build vs. buy analysis: Compare managed API costs against self-hosting open-weight models (Llama, DeepSeek, Mistral) on your own infrastructure.
  • Flat-rate evaluation: For teams with 50M+ tokens/month, Plugsky's flat-rate plans often beat per-token pricing. Use the cost calculator to model your scenario.

Limitations

  • Prices change frequently. This index reflects a single point in time (July 11, 2026). Always verify on the provider's official pricing page.
  • Volume discounts not included. Enterprise contracts for OpenAI, Anthropic, Google, and others can reduce per-token pricing by 20-50% at scale.
  • Batch vs. real-time pricing. Some providers (OpenAI, Anthropic) offer lower batch/async pricing (50% discount). This table shows standard real-time rates.
  • Cached pricing. Providers like Anthropic and Google reduce pricing for previously-cached context. Standard rates shown.
  • Regional variation. Prices may differ by region (US, EU, Asia). US-based pricing shown.
  • Plugsky comparison note. Plugsky flat-rate pricing is not directly comparable to per-token pricing. Evaluate based on total monthly spend, not per-token cost.

Changelog

  • July 11, 2026 — Initial publication. 31 models across 11 providers.

Frequently asked questions

Which AI API is cheapest?

On a per-token basis, Google Gemini 2.0 Flash ($0.10/$0.40 per MTok) and DeepSeek-Coder-V4 ($0.28/$1.10 per MTok) are the cheapest listed providers. However, Plugsky's flat-rate plans ($20-$120/month for unlimited tokens) offer the best value for teams with moderate to high usage — the per-token cost approaches zero at scale.

Does Plugsky charge per token?

No. Plugsky self-serve plans are flat monthly — $20 (Starter), $60 (Builder), or $120 (Scale) for unlimited usage on every model in your tier. There is no per-token metering on self-serve plans. Enterprise plans are annual contracts with custom rate limits. See full Plugsky pricing details.

How often is this pricing index updated?

This index is updated monthly. The current edition was published July 11, 2026. Prices change frequently — always verify on the provider's official pricing page before making purchasing decisions. Check the changelog for update history.

Can I download this data?

Yes. The full pricing table above can be copied directly as CSV. Select the table, copy, and paste into any spreadsheet. A formatted CSV block is also provided in the Download section. The dataset is released under a CC0 public domain dedication — use freely.

Which provider has the best value?

Value depends on your usage pattern. For light usage (<10M tokens/month), per-token APIs from Google or DeepSeek are cheapest. For moderate to heavy usage (50M+ tokens/month), Plugsky's flat-rate plans eliminate cost uncertainty. Frontier intelligence vs. cost decisions should also consider context window size, speed, and accuracy benchmarks. Use the LLM cost calculator to model your specific workload.

Start building with Plugsky

Free trial with plugsky-micro. No card required. Then Starter is $20/mo flat — unlimited tokens.

Start Free → Plugsky pricing

Related

AI API pricing

Plugsky flat-rate pricing explained: $20, $60, $120 for unlimited usage.

OpenAI-compatible API

Drop-in replacement for OpenAI with 31+ models and flat-rate pricing.

LLM cost calculator

Compare per-token vs. flat-rate pricing for your workload.