Data updated November 21, 2026

Use AI,
pay less.

AI model prices in one place: what each vendor charges, who sells the same model for less, which modes cut the bill in half, and what building a project would cost. Every number has a source and a date.

Cheapest current model of each vendor

per 1M tokens

  1. Nemotron 3.5 Lightning NVIDIA · via DeepInfra$0.085
  2. DeepSeek V4 Flash 0731 DeepSeek · via DeepInfra$0.09
  3. Muse Spark 1.3 Contributor Meta$0.125
  4. Claude Haiku 5.5 Anthropic$0.20
  5. Qwen3.8 Omni Flash Qwen (Alibaba)$0.23
  6. GLM 5.3 Flash Z.AI$0.2375

Mixed price: three input tokens to one output token, per 1M tokens. Cheaper does not mean equally capable, so compare models of the same class.

Models with a price
341
Price confirmed by two sources
96
Free models
12

For 1 USD you get about 0.58 million tokens at a typical price of $1.71 per million.

The currency applies to prices, the calculator and the table; plan prices and the chart stay in US dollars. Exchange rate, NBP: November 20, 2026

Cheapest right now

Biggest savings at another provider

  1. DeepSeek V4 Pro 0813 via Baidu −90% promo fp8
  2. DeepSeek V4.1 Flash 1017 via Sail Research −90% fp4
  3. DeepSeek V4 Flash 0423 via Baidu −84% promo fp8
  4. GLM 5.3 Flash via Relace −79% fp4
  5. GLM 5.2 via StreamLake −78% promo fp8

The same model sold by a different provider. Check the quantization and any provider discount (promo) before you rely on the price.

Same model, smaller bill

  • Cached input tokens cost 90% less (median of current models).
  • Batch mode: 50% off (models: 70).
  • Flex tier: 50% off (current models: 8), in exchange for slower or less predictable answers.
  • Models with a free tier: 12.

What building a project costs

Pick a task size. The prices are real; how much text a task needs is our assumption, so move the sliders to match your case.

Total: 1,200,000 input and 60,000 output tokens.

From $0.05 to $7.07 for this task, depending on the model.

  1. DeepSeek V4 Flash 0731 DeepSeek · via DeepInfra$0.05
  2. Muse Spark 1.3 Contributor Meta$0.05×1.1
  3. Nemotron 3.5 Lightning NVIDIA · via DeepInfra$0.06×1.2
  4. Muse Ember Meta$0.07×1.6
  5. DeepSeek V4.1 Flash DeepSeek · via River$0.08×1.7
  6. Claude Haiku 5.5 Anthropic$0.08×1.7
  7. Qwen3.8 Omni Flash Qwen (Alibaba)$0.10×2.1
  8. Qwen3.8 Flash Qwen (Alibaba)$0.10×2.2
  9. Mistral Small 4 Mistral AI$0.11×2.3
  10. MiniMax M3.1 Flash MiniMax$0.11×2.3
  11. Models not shown: 48
  12. Claude Fable 5.2 Anthropic$5.65×123
  13. GPT-6.2 Astra Pro OpenAI$5.82×127
  14. GPT-6.2 Astra OpenAI$5.82×127
  15. Claude Fable 5.1 Anthropic$7.07×154

Bars use a logarithmic scale so cheap and expensive models fit on one chart: a bar twice as long is many times the price. Compare the amounts on the right.

An estimate, not a measurement: the number of calls and the tokens per call are assumptions, and caching is modelled as one cache write followed by cache reads. Real projects vary a lot.

Compare models

Model Inputper 1M tokens Outputper 1M tokens Cachecached input Context Cheaper elsewhereempty: no cheaper provider found Cheaper modes Data
Nemotron 3.5 Lightning NVIDIA Input per 1M tokens$0.06via DeepInfra Output per 1M tokens$0.16 Cache cached input−50% Context1M Cheaper elsewhere Darkbloom $0.02 / $0.096 −54% int4 Cheaper modes — Data one source
DeepSeek V4 Flash 0731 DeepSeek Input per 1M tokens$0.06via DeepInfra Output per 1M tokens$0.18 Cache cached input−75% Context1M Cheaper elsewhere Baidu $0.044 / $0.132 −27% promo −90% fp8 Cheaper modes flex −20% Data disputed
Nemotron 3 Super NVIDIA Input per 1M tokens$0.075via DeepInfra Output per 1M tokens$0.40 Cache cached input— Context262k Cheaper elsewhere — Cheaper modes flex −20% Data one source
Muse Spark 1.3 Contributor Meta Input per 1M tokens$0.10 Output per 1M tokens$0.20 Cache cached input−98% Context1M Cheaper elsewhere — Cheaper modes — Data one source
Nemotron 3.5 Super VL NVIDIA Input per 1M tokens$0.10via Crusoe Output per 1M tokens$0.40 Cache cached input−20% Context1M Cheaper elsewhere Darkbloom $0.06 / $0.35 −24% nvfp4 Cheaper modes — Data one source
Claude Haiku 5.5 Anthropic Input per 1M tokens$0.10above 100k: $0.50 Output per 1M tokens$0.50 Cache cached input−90% Context1M Cheaper elsewhere — Cheaper modes batch −50% Data confirmed
Muse Ember Meta Input per 1M tokens$0.10 Output per 1M tokens$0.50 Cache cached input−96% Context1M Cheaper elsewhere — Cheaper modes — Data one source
DeepSeek V4.1 Flash DeepSeek Input per 1M tokens$0.12via River Output per 1M tokens$0.48 Cache cached input−98% Context1M Cheaper elsewhere Baidu $0.0507 / $0.2028 −58% promo −83% fp8 Cheaper modes — Data disputed
Qwen3.8 Flash Qwen (Alibaba) Input per 1M tokens$0.15 Output per 1M tokens$0.47 Cache cached input−89% Context1M Cheaper elsewhere — Cheaper modes — Data one source
Qwen3.8 Omni Flash Qwen (Alibaba) Input per 1M tokens$0.15 Output per 1M tokens$0.47 Cache cached input−89% Context1M Cheaper elsewhere — Cheaper modes — Data one source
GLM 5.3 Flash Z.AI Input per 1M tokens$0.15 Output per 1M tokens$0.50 Cache cached input−80% Context1M Cheaper elsewhere Relace $0.0399 / $0.0784 −79% fp4 Cheaper modes — Data one source
MiniMax M3.1 Flash MiniMax Input per 1M tokens$0.15time-of-day price Output per 1M tokens$0.60 Cache cached input−90% Context1M Cheaper elsewhere Makora $0.17 / $0.40 −13% fp4 Cheaper modes — Data one source
Mistral Small 4 Mistral AI Input per 1M tokens$0.15 Output per 1M tokens$0.60 Cache cached input−90% Context262k Cheaper elsewhere — Cheaper modes — Data confirmed
Nemotron 3.5 Content Safety NVIDIA Input per 1M tokens$0.20via DeepInfra Output per 1M tokens$0.20 Cache cached input— Context131k Cheaper elsewhere — Cheaper modes — Data one source
MiniMax M2.5 MiniMax Input per 1M tokens$0.27via Venice Output per 1M tokens$0.95 Cache cached input−89% Context198k Cheaper elsewhere — Cheaper modes — Data one source
DeepSeek V4.1 Flash 1017 DeepSeek Input per 1M tokens$0.30time-of-day price Output per 1M tokens$1.20 Cache cached input−98% Context1M Cheaper elsewhere Sail Research $0.04 / $0.10 −90% fp4 Cheaper modes — Data one source
MiniMax M3 MiniMax Input per 1M tokens$0.30 Output per 1M tokens$1.20 Cache cached input−80% Context1M Cheaper elsewhere CoreWeave $0.23 / $0.96 −21% fp4 Cheaper modes — Data one source
Command A+ Cohere Input per 1M tokens$0.30 Output per 1M tokens$1.50 Cache cached input−50% Context192k Cheaper elsewhere — Cheaper modes — Data one source
Nova 2 Lite Amazon Input per 1M tokens$0.30 Output per 1M tokens$2.50 Cache cached input— Context1M Cheaper elsewhere — Cheaper modes — Data one source
GLM 5.3 FlashX Z.AI Input per 1M tokens$0.37 Output per 1M tokens$1.25 Cache cached input−80% Context1M Cheaper elsewhere — Cheaper modes — Data one source
GLM 5.3 FlashX Legal Z.AI Input per 1M tokens$0.37 Output per 1M tokens$1.25 Cache cached input−80% Context1M Cheaper elsewhere — Cheaper modes — Data one source
Devstral 2 2512 Mistral AI Input per 1M tokens$0.40 Output per 1M tokens$2.00 Cache cached input−90% Context262k Cheaper elsewhere — Cheaper modes — Data one source
DeepSeek V4 Flash Vision Exp DeepSeek Input per 1M tokens$0.44via GMICloud Output per 1M tokens$1.32 Cache cached input−97% Context1M Cheaper elsewhere DeepInfra $0.2156 / $0.6468 −51% promo −51% fp8 Cheaper modes — Data one source
Kimi K2.5 Moonshot AI Input per 1M tokens$0.49via AtlasCloudint4 Output per 1M tokens$2.50 Cache cached input−59% Context262k Cheaper elsewhere — Cheaper modes — Data one source
Nemotron 3 Ultra NVIDIA Input per 1M tokens$0.50via KT Output per 1M tokens$2.20 Cache cached input−80% Context262k Cheaper elsewhere — Cheaper modes — Data one source
MiniMax M2.7 MiniMax Input per 1M tokens$0.60via Groq Output per 1M tokens$1.80 Cache cached input— Context197k Cheaper elsewhere GMICloud $0.21 / $0.84 −59% promo −30% fp8 Cheaper modes — Data one source
Mistral Large 4 Mistral AI Input per 1M tokens$0.70 Output per 1M tokens$2.10 Cache cached input−90% Context1M Cheaper elsewhere — Cheaper modes — Data confirmed
Gemini 3.8 Flash Google Input per 1M tokens$0.75promo −50% Output per 1M tokens$3.75 Cache cached input−90% Context1M Cheaper elsewhere — Cheaper modes flex −50% batch −50% Data confirmed
Kimi K2.6 Moonshot AI Input per 1M tokens$0.95 Output per 1M tokens$4.00 Cache cached input−83% Context262k Cheaper elsewhere Baidu $0.3895 / $1.64 −59% promo −59% fp4 Cheaper modes — Data one source
Kimi K2.7 Code Moonshot AI Input per 1M tokens$0.95 Output per 1M tokens$4.00 Cache cached input−80% Context262k Cheaper elsewhere CoreWeave $0.60 / $3.00 −30% int4 Cheaper modes — Data one source
GPT-6.1 Terra Pro OpenAI Input per 1M tokens$1.00above 272k: $2.00 Output per 1M tokens$5.00 Cache cached input−95% Context1.1M Cheaper elsewhere — Cheaper modes flex −50% batch −50% Data one source
Grok 4.20 Multi-Agent xAI Input per 1M tokens$1.25above 200k: $2.50 Output per 1M tokens$2.50 Cache cached input−84% Context2M Cheaper elsewhere — Cheaper modes — Data confirmed
Grok 4.3 xAI Input per 1M tokens$1.25above 200k: $2.50 Output per 1M tokens$2.50 Cache cached input−84% Context1M Cheaper elsewhere — Cheaper modes — Data confirmed
Muse Spark 1.3 Meta Input per 1M tokens$1.25 Output per 1M tokens$4.25 Cache cached input−88% Context1M Cheaper elsewhere — Cheaper modes — Data one source
Muse Spark Open Meta Input per 1M tokens$1.25via Fireworks Output per 1M tokens$4.25 Cache cached input−88% Context1M Cheaper elsewhere — Cheaper modes — Data one source
DeepSeek V4 Pro 0813 DeepSeek Input per 1M tokens$1.32time-of-day price Output per 1M tokens$3.96 Cache cached input−97% Context1M Cheaper elsewhere Baidu $0.132 / $0.396 −90% promo −90% fp8 Cheaper modes — Data disputed
GLM 5.3 Z.AI Input per 1M tokens$1.40 Output per 1M tokens$4.40 Cache cached input−81% Context1M Cheaper elsewhere Wafer $0.179 / $1.40 −78% fp4 Cheaper modes — Data one source
Qwen4 Max Qwen (Alibaba) Input per 1M tokens$1.43promo −5% Output per 1M tokens$4.28 Cache cached input−95% Context1M Cheaper elsewhere — Cheaper modes — Data one source
Gemini 4 Hydrogen Google Input per 1M tokens$1.50 Output per 1M tokens$7.50 Cache cached input−95% Context1M Cheaper elsewhere — Cheaper modes flex −50% batch −50% Data confirmed
Mistral Medium 3.5 Mistral AI Input per 1M tokens$1.50 Output per 1M tokens$7.50 Cache cached input— Context262k Cheaper elsewhere — Cheaper modes batch −50% Data confirmed
GLM 5.3 Prime Z.AI Input per 1M tokens$1.68via Alibaba Output per 1M tokens$5.28 Cache cached input−80% Context1M Cheaper elsewhere — Cheaper modes — Data one source
Grok 4.5 xAI Input per 1M tokens$2.00above 200k: $4.00 Output per 1M tokens$6.00 Cache cached input−85% Context500k Cheaper elsewhere — Cheaper modes — Data confirmed
Grok 4.6 xAI Input per 1M tokens$2.00above 200k: $4.00 Output per 1M tokens$6.00 Cache cached input−75% Context500k Cheaper elsewhere — Cheaper modes — Data confirmed
Grok 4.7 xAI Input per 1M tokens$2.00above 200k: $4.00 Output per 1M tokens$6.00 Cache cached input−75% Context500k Cheaper elsewhere — Cheaper modes — Data confirmed
Pixtral Large 2502 Mistral AI Input per 1M tokens$2.00via Amazon Bedrock Output per 1M tokens$6.00 Cache cached input— Context128k Cheaper elsewhere — Cheaper modes — Data one source
Qwen3.8 Max (0902) Qwen (Alibaba) Input per 1M tokens$2.00 Output per 1M tokens$6.00 Cache cached input−88% Context1M Cheaper elsewhere — Cheaper modes batch −43% Data one source
Claude Sonnet 5.5 Anthropic Input per 1M tokens$2.00 Output per 1M tokens$10.00 Cache cached input−95% Context1M Cheaper elsewhere — Cheaper modes batch −50% Data confirmed
Gemini 4 Argon Google Input per 1M tokens$2.00promo −50% Output per 1M tokens$10.00 Cache cached input−95% Context1M Cheaper elsewhere — Cheaper modes flex −50% batch −50% Data confirmed
Gemini 4 Carbon Google Input per 1M tokens$2.00promo −50% Output per 1M tokens$10.00 Cache cached input−95% Context1M Cheaper elsewhere — Cheaper modes flex −50% batch −50% Data confirmed
GPT-6.2 Sol OpenAI Input per 1M tokens$2.00above 272k: $4.00 Output per 1M tokens$10.00 Cache cached input−95% Context1.1M Cheaper elsewhere — Cheaper modes batch −50% Data confirmed
GPT-6.2 Sol Pro OpenAI Input per 1M tokens$2.00above 272k: $4.00 Output per 1M tokens$10.00 Cache cached input−95% Context1.1M Cheaper elsewhere — Cheaper modes batch −50% Data one source
Kimi K3.1 Pro Moonshot AI Input per 1M tokens$2.00mxfp4 Output per 1M tokens$10.00 Cache cached input−90% Context1M Cheaper elsewhere Relace $0.2403 / $8.40 −43% fp4 Cheaper modes batch −58% Data one source
Gemini 4 Neon Google Input per 1M tokens$3.00 Output per 1M tokens$15.00 Cache cached input−95% Context1M Cheaper elsewhere — Cheaper modes flex −50% batch −50% Data confirmed
Kimi K3 Moonshot AI Input per 1M tokens$3.00 Output per 1M tokens$15.00 Cache cached input−90% Context1M Cheaper elsewhere Relace $0.1805 / $11.01 −52% fp4 Cheaper modes — Data one source
Sonar Pro Search Perplexity Input per 1M tokens$3.00 Output per 1M tokens$15.00 Cache cached input— Context200k Cheaper elsewhere — Cheaper modes — Data one source
Qwen3.8 Max Prime Qwen (Alibaba) Input per 1M tokens$4.00 Output per 1M tokens$12.00 Cache cached input−88% Context1M Cheaper elsewhere — Cheaper modes — Data one source
Claude Opus 5.5 Anthropic Input per 1M tokens$4.00 Output per 1M tokens$20.00 Cache cached input−95% Context1M Cheaper elsewhere — Cheaper modes batch −50% Data confirmed
Muse Wonder Meta Input per 1M tokens$4.00 Output per 1M tokens$20.00 Cache cached input−95% Context1M Cheaper elsewhere — Cheaper modes — Data one source
Claude Fable 5.2 Anthropic Input per 1M tokens$8.00 Output per 1M tokens$40.00 Cache cached input−98% Context1M Cheaper elsewhere — Cheaper modes batch −50% Data confirmed
GPT-6.2 Astra OpenAI Input per 1M tokens$8.00above 272k: $16.00 Output per 1M tokens$40.00 Cache cached input−95% Context1.1M Cheaper elsewhere — Cheaper modes batch −50% Data confirmed
GPT-6.2 Astra Pro OpenAI Input per 1M tokens$8.00above 272k: $16.00 Output per 1M tokens$40.00 Cache cached input−95% Context1.1M Cheaper elsewhere — Cheaper modes batch −50% Data one source
Claude Fable 5.1 Anthropic Input per 1M tokens$10.00 Output per 1M tokens$50.00 Cache cached input−98% Context1M Cheaper elsewhere — Cheaper modes batch −50% Data confirmed

Prices are USD per million tokens converted at the NBP rate. Where the vendor sells no API of its own (open-weight models), the cheapest hosting provider is shown and named.

Price ranges by vendor

One row per vendor, one dot per current model, on a logarithmic price axis: each step to the right is ten times the price. The yellow dot is the vendor's cheapest model and is named. Hover or tap a dot for its price.

0.10.1111010Mixed price, USD per 1M tokensNVIDIANemotron 3 Ultra (NVIDIA · via KT): $0.925Nemotron 3.5 Content Safety (NVIDIA · via DeepInfra): $0.20Nemotron 3.5 Super VL (NVIDIA · via Crusoe): $0.175Nemotron 3 Super (NVIDIA · via DeepInfra): $0.1563Nemotron 3.5 Lightning (NVIDIA · via DeepInfra): $0.085Nemotron 3.5 Lightning · via DeepInfraDeepSeekDeepSeek V4 Pro 0813 (DeepSeek): $1.98DeepSeek V4 Flash Vision Exp (DeepSeek · via GMICloud): $0.66DeepSeek V4.1 Flash 1017 (DeepSeek): $0.525DeepSeek V4.1 Flash (DeepSeek · via River): $0.21DeepSeek V4 Flash 0731 (DeepSeek · via DeepInfra): $0.09DeepSeek V4 Flash 0731 · via DeepInfraMetaMuse Wonder (Meta): $8.00Muse Spark 1.3 (Meta): $2.00Muse Spark Open (Meta · via Fireworks): $2.00Muse Ember (Meta): $0.20Muse Spark 1.3 Contributor (Meta): $0.125Muse Spark 1.3 ContributorAnthropicClaude Fable 5.1 (Anthropic): $20.00Claude Fable 5.2 (Anthropic): $16.00Claude Opus 5.5 (Anthropic): $8.00Claude Sonnet 5.5 (Anthropic): $4.00Claude Haiku 5.5 (Anthropic): $0.20Claude Haiku 5.5Qwen (Alibaba)Qwen3.8 Max Prime (Qwen (Alibaba)): $6.00Qwen3.8 Max (0902) (Qwen (Alibaba)): $3.00Qwen4 Max (Qwen (Alibaba)): $2.14Qwen3.8 Flash (Qwen (Alibaba)): $0.23Qwen3.8 Omni Flash (Qwen (Alibaba)): $0.23Qwen3.8 Omni FlashZ.AIGLM 5.3 Prime (Z.AI · via Alibaba): $2.58GLM 5.3 (Z.AI): $2.15GLM 5.3 FlashX (Z.AI): $0.59GLM 5.3 FlashX Legal (Z.AI): $0.59GLM 5.3 Flash (Z.AI): $0.2375GLM 5.3 FlashMistral AIMistral Medium 3.5 (Mistral AI): $3.00Pixtral Large 2502 (Mistral AI · via Amazon Bedrock): $3.00Mistral Large 4 (Mistral AI): $1.05Devstral 2 2512 (Mistral AI): $0.80Mistral Small 4 (Mistral AI): $0.2625Mistral Small 4MiniMaxMiniMax M2.7 (MiniMax · via Groq): $0.90MiniMax M3 (MiniMax): $0.525MiniMax M2.5 (MiniMax · via Venice): $0.44MiniMax M3.1 Flash (MiniMax): $0.2625MiniMax M3.1 FlashCohereCommand A+ (Cohere): $0.60Command A+AmazonNova 2 Lite (Amazon): $0.85Nova 2 LiteMoonshot AIKimi K3 (Moonshot AI): $6.00Kimi K3.1 Pro (Moonshot AI): $4.00Kimi K2.6 (Moonshot AI): $1.71Kimi K2.7 Code (Moonshot AI): $1.71Kimi K2.5 (Moonshot AI · via AtlasCloud): $0.9925Kimi K2.5 · via AtlasCloudGoogleGemini 4 Neon (Google): $6.00Gemini 4 Argon (Google): $4.00Gemini 4 Carbon (Google): $4.00Gemini 4 Hydrogen (Google): $3.00Gemini 3.8 Flash (Google): $1.50Gemini 3.8 FlashxAIGrok 4.5 (xAI): $3.00Grok 4.6 (xAI): $3.00Grok 4.7 (xAI): $3.00Grok 4.20 Multi-Agent (xAI): $1.56Grok 4.3 (xAI): $1.56Grok 4.3OpenAIGPT-6.2 Astra (OpenAI): $16.00GPT-6.2 Astra Pro (OpenAI): $16.00GPT-6.2 Sol (OpenAI): $4.00GPT-6.2 Sol Pro (OpenAI): $4.00GPT-6.1 Terra Pro (OpenAI): $2.00GPT-6.1 Terra ProPerplexitySonar Pro Search (Perplexity): $6.00Sonar Pro Search
0.10.1111010Mixed price, USD per 1M tokensNVIDIANemotron 3 Ultra (NVIDIA · via KT): $0.925Nemotron 3.5 Content Safety (NVIDIA · via DeepInfra): $0.20Nemotron 3.5 Super VL (NVIDIA · via Crusoe): $0.175Nemotron 3 Super (NVIDIA · via DeepInfra): $0.1563Nemotron 3.5 Lightning (NVIDIA · via DeepInfra): $0.085Nemotron 3.5 Lightning · via DeepInfraDeepSeekDeepSeek V4 Pro 0813 (DeepSeek): $1.98DeepSeek V4 Flash Vision Exp (DeepSeek · via GMICloud): $0.66DeepSeek V4.1 Flash 1017 (DeepSeek): $0.525DeepSeek V4.1 Flash (DeepSeek · via River): $0.21DeepSeek V4 Flash 0731 (DeepSeek · via DeepInfra): $0.09DeepSeek V4 Flash 0731 · via DeepInfraMetaMuse Wonder (Meta): $8.00Muse Spark 1.3 (Meta): $2.00Muse Spark Open (Meta · via Fireworks): $2.00Muse Ember (Meta): $0.20Muse Spark 1.3 Contributor (Meta): $0.125Muse Spark 1.3 ContributorAnthropicClaude Fable 5.1 (Anthropic): $20.00Claude Fable 5.2 (Anthropic): $16.00Claude Opus 5.5 (Anthropic): $8.00Claude Sonnet 5.5 (Anthropic): $4.00Claude Haiku 5.5 (Anthropic): $0.20Claude Haiku 5.5Qwen (Alibaba)Qwen3.8 Max Prime (Qwen (Alibaba)): $6.00Qwen3.8 Max (0902) (Qwen (Alibaba)): $3.00Qwen4 Max (Qwen (Alibaba)): $2.14Qwen3.8 Flash (Qwen (Alibaba)): $0.23Qwen3.8 Omni Flash (Qwen (Alibaba)): $0.23Qwen3.8 Omni FlashZ.AIGLM 5.3 Prime (Z.AI · via Alibaba): $2.58GLM 5.3 (Z.AI): $2.15GLM 5.3 FlashX (Z.AI): $0.59GLM 5.3 FlashX Legal (Z.AI): $0.59GLM 5.3 Flash (Z.AI): $0.2375GLM 5.3 FlashMistral AIMistral Medium 3.5 (Mistral AI): $3.00Pixtral Large 2502 (Mistral AI · via Amazon Bedrock): $3.00Mistral Large 4 (Mistral AI): $1.05Devstral 2 2512 (Mistral AI): $0.80Mistral Small 4 (Mistral AI): $0.2625Mistral Small 4MiniMaxMiniMax M2.7 (MiniMax · via Groq): $0.90MiniMax M3 (MiniMax): $0.525MiniMax M2.5 (MiniMax · via Venice): $0.44MiniMax M3.1 Flash (MiniMax): $0.2625MiniMax M3.1 FlashCohereCommand A+ (Cohere): $0.60Command A+AmazonNova 2 Lite (Amazon): $0.85Nova 2 LiteMoonshot AIKimi K3 (Moonshot AI): $6.00Kimi K3.1 Pro (Moonshot AI): $4.00Kimi K2.6 (Moonshot AI): $1.71Kimi K2.7 Code (Moonshot AI): $1.71Kimi K2.5 (Moonshot AI · via AtlasCloud): $0.9925Kimi K2.5 · via AtlasCloudGoogleGemini 4 Neon (Google): $6.00Gemini 4 Argon (Google): $4.00Gemini 4 Carbon (Google): $4.00Gemini 4 Hydrogen (Google): $3.00Gemini 3.8 Flash (Google): $1.50Gemini 3.8 FlashxAIGrok 4.5 (xAI): $3.00Grok 4.6 (xAI): $3.00Grok 4.7 (xAI): $3.00Grok 4.20 Multi-Agent (xAI): $1.56Grok 4.3 (xAI): $1.56Grok 4.3OpenAIGPT-6.2 Astra (OpenAI): $16.00GPT-6.2 Astra Pro (OpenAI): $16.00GPT-6.2 Sol (OpenAI): $4.00GPT-6.2 Sol Pro (OpenAI): $4.00GPT-6.1 Terra Pro (OpenAI): $2.00GPT-6.1 Terra ProPerplexitySonar Pro Search (Perplexity): $6.00Sonar Pro Search

Speed is not shown yet. No public source publishes it, so we measure it ourselves and the first measurements will appear here.

Subscriptions and plans

Consumer plans as the vendors list them on their own pages. Regional prices and taxes may differ. Prices are in US dollars exactly as the vendors list them on their US pages. They are not converted: prices in other countries differ, and so do taxes.

checked October 22, 2026

Show all 18

Discounts and promotions

Discounts inside the price

These prices already include a discount declared by the provider. We do not know when it ends.

Free tiers, credits and programs

Titles and details are quoted from the providers' own pages and stay in English. Each line links to its source and shows the date we checked it. Check the provider's page before you rely on an offer.

checked October 22, 2026

Discounts 11

Show all 11

Free tiers 11

Show all 11

Credits 8

Show all 8

Student and teacher offers 12

Show all 12

Startup programs 12

Show all 12

Promo codes for AI services submitted by users are collected on kodyrabatowe.cleanwhale.pl.

How to pay less

  1. Reuse long prompts (prompt caching)

    Cached input is billed at a fraction of the normal price: the median saving is 90% across 55 current models. Put the part of the prompt that never changes first.

  2. Use batch or flex when you can wait

    Batch and flex modes cut the price by about 50%, in exchange for slower or less predictable answers.

  3. Use the small model for simple work

    Within one vendor the cheapest current model is typically 10.9 times cheaper than the dearest. Example: Claude Haiku 5.5 costs 1/100 of Claude Fable 5.1 (Anthropic).

  4. Compare hosting providers

    The same model can cost very different amounts. Llama 3.1 8B Instruct: 4 providers, from DeepInfra to CoreWeave, a 8.8x spread.

  5. Keep answers short

    Output tokens usually cost about 4 times more than input tokens, so asking for concise answers saves more than shortening the question.

  6. Mind the region

    Regional endpoints (EU or US) at Mistral, Alibaba, and Fireworks cost about 10% more than the global route.

Method and sources

Where the numbers come from
Prices and routes: OpenRouter API. Second opinion on vendor prices: LiteLLM. Exchange rates: NBP (official mid rates, published on business days).
What "confirmed" means
The price the vendor itself charges is read from the vendor's own route and compared with an independent price table. Only a match within half a percent is marked confirmed. Models with a single source are shown, labelled, and never presented as verified.
Cheaper elsewhere
For every model we list each provider's ordinary route. "Cheaper elsewhere" is the lowest of them against the vendor's own price. Flex, priority, regional and special variants are kept apart and never counted as the ordinary price.
The cost calculator
Prices come from the table. The size of the task is an assumption: calls times tokens per call, with a cached share. Treat the result as an order of magnitude and set the sliders to your case.
Speed
No public source publishes provider speed, so it is measured by our own test calls. Until the first measurements exist, the column is hidden instead of showing guesses.
Freshness
Prices refreshed November 21, 2026; exchange rate of November 20, 2026. Prices change often, so check the vendor's page before you pay.
Independence
We do not rank models by quality and we do not recommend a vendor.