Skip to content
AI Dev Toolkit.
Esc
  • Loading…

Tokens & Costs

AI model comparison: LLM pricing, context windows and limits

Compare 340+ AI models side by side on what they cost through the API: input and output price per million tokens, context window, maximum output, caching and release date, in one sortable table updated daily from public OpenRouter and LiteLLM data (prices as of 22 November 2026). It compares price and specs, not benchmark scores.

Providers
Must support

Showing 50 of 324 matching models (324 in total)

Prices updated 2026-11-22
AI model prices in US dollars per million tokens, context windows and release dates. Column headers sort the table.
Cost estimate
DaisyRose 27BRadixArkOpen$0.15$1.875$0.0375262K–$0.03932026-11-21Estimate cost
Inkling 2 MiniThinking MachinesOpen$0.075$0.30$0.0151.05M66K$0.07862026-11-21Estimate cost
Pixtral Large 2502Mistral$2$6–128K–$0.262026-11-20Estimate cost
Gemini 4 NeonGoogle$3$15$0.151.05M–$3.152026-11-18Estimate cost
GPT-6.2 AstraOpenAI$8$40$0.401.05M128K$16.802026-11-17Estimate cost
GPT-6.2 Astra ProOpenAI$8$40$0.401.05M128K$16.802026-11-17Estimate cost
GPT-6.2 SolOpenAI$2$10$0.101.05M128K$4.202026-11-17Estimate cost
GPT-6.2 Sol ProOpenAI$2$10$0.101.05M128K$4.202026-11-17Estimate cost
Hy4 flashTencentOpen$0.035$0.1405$0.00151.05M66K$0.03672026-11-16Estimate cost
Qwen4 MaxQwen$1.425$4.275$0.07131M131K$1.432026-11-16Estimate cost
Gemini 4 HydrogenGoogle$1.50$7.50$0.0751.05M–$1.572026-11-12Estimate cost
ERNIE-Light-DT-1.0Baidu$1.13$3.94$0.1131.02M131K$1.162026-11-10Estimate cost
Gemini 4 CarbonGoogle$2$10$0.101.05M–$2.102026-11-10Estimate cost
Nex-N2.6Nex AGIOpen$0.12$0.40$0.024262K–$0.03152026-11-10Estimate cost
Claude Fable 5.2Anthropic$8$40$0.201M128K$8.002026-11-09Estimate cost
Inkling 2 SmallThinking MachinesOpen$0.30$1$0.031.05M–$0.312026-11-09Estimate cost
Muse EmberMeta$0.10$0.50$0.0041.05M–$0.102026-11-09Estimate cost
LFM3-1.7BLiquidAIOpen$0.08$0.20$0.02262K8K$0.0212026-11-06Estimate cost
Mi:dm K 3.0 ProKT$0.20$0.80$0.04262K–$0.05242026-11-05Estimate cost
Hy4TencentOpen$0.417$1.2505$0.0211.05M66K$0.442026-11-04Estimate cost
Ripple 65B PreviewRicursiveOpen$0.10$0.30$0.0251.05M131K$0.102026-11-04Estimate cost
GLM 5.3 FlashX LegalZ.ai$0.37$1.25$0.0751.05M131K$0.392026-11-02Estimate cost
Motif 3Motif TechnologiesOpen$0.30$1.20$0.06262K–$0.07862026-11-02Estimate cost
Particle-1Particle AI$0.07$0.50$0.031.05M131K$0.07342026-10-28Estimate cost
GPT-6.1 TerraOpenAI$1$5$0.051.05M128K$2.102026-10-26Estimate cost
GPT-6.1 Terra ProOpenAI$1$5$0.051.05M128K$2.102026-10-26Estimate cost
Kimi K3.1 ProMoonshot AIOpen$2$10$0.201.05M–$2.102026-10-23Estimate cost
MiniMax M3.1 FlashMiniMaxOpen$0.15$0.60$0.0151.05M524K$0.162026-10-19Estimate cost
DeepSeek V4.1 Flash 1017DeepSeekOpen$0.30$1.20$0.0061.05M393K$0.312026-10-17Estimate cost
GlimpseElorian$0.60$3$0.06262K–$0.162026-10-17Estimate cost
Beam 1.0 501B A23BReflectionOpen$0.25$1$0.051.05M131K$0.262026-10-16Estimate cost
Ternary Bonsai 2 9BPrismMLOpen$0.08$0.13$0.04262K66K$0.0212026-10-16Estimate cost
Moondream 3 PreviewMoondreamOpen$0.30$2.50–32K4K$0.00962026-10-15Estimate cost
Moondream 3.1MoondreamOpen$0.30$2.50–32K4K$0.00962026-10-15Estimate cost
Muse Spark OpenMetaOpen$1.25$4.25$0.151.05M–$1.312026-10-15Estimate cost
Nemotron 3.5 Super VLNVIDIAOpen$0.15$0.47$0.021.05M–$0.162026-10-15Estimate cost
Pareto 26.10unbiased$0.80$3.20$0.031.05M131K$0.842026-10-15Estimate cost
ExtractScaleDown$0.04$0–1M–$0.042026-10-13Estimate cost
Gemini 4 ArgonGoogle$2$10$0.101.05M–$2.102026-10-13Estimate cost
LFM2.5-24B-A2BLiquidAIOpen$0.05$0.70$0.025128K8K$0.00642026-10-13Estimate cost
Pareto 26.9unbiased$2.50$7.50$0.25262K131K$0.662026-10-13Estimate cost
SummarizeScaleDown$0.04$0–1M4K$0.042026-10-13Estimate cost
Muse WonderMeta$4$20$0.201.05M–$4.192026-10-12Estimate cost
Step 5 PreviewStepFunOpen$1$2.70$0.101M64K$1.002026-10-08Estimate cost
Claude Haiku 5.5Anthropic$0.10$0.50$0.011M128K$0.502026-10-07Estimate cost
Mistral Large 4MistralOpen$0.70$2.10$0.071.05M262K$0.732026-10-06Estimate cost
Ling 3.1 FlashinclusionAIOpen$0.06$0.18$0.0121M262K$0.062026-10-02Estimate cost
Pareto 26.10 Previewunbiased$0.80$3.20$0.031.05M131K$0.842026-10-01Estimate cost
GPT-6.1 SolOpenAI$2$10$0.101.05M128K$4.202026-09-29Estimate cost
Claude Sonnet 5.5Anthropic$2$10$0.101M128K$2.002026-09-28Estimate cost

Steps

How do I compare AI models side by side?

  1. Search for a model, or pick one or more providers.
  2. Set a minimum context window and any capabilities you need, such as image input or tool calling.
  3. Click a column header to sort, for example by input price or context window. Click again to reverse.
  4. Use “Estimate cost” on any row to price your own workload in the LLM cost calculator.
  5. Copy the page address to share the exact filters and sort you’re looking at.

Method

How do I read an LLM pricing comparison?

Choosing a model usually starts with three questions: can it handle my input, can it do what I need, and what will it cost? This table answers all three for 340+ models at once, so you can narrow the field before testing anything. It doesn’t rank quality: the data has no benchmark scores, so pair it with a leaderboard if you need one.

Which AI model is the cheapest?

On input price, Ling 3.0 Tiny from inclusionAI: $0.015 per million input tokens and $0.045 per million output tokens (as of 22 November 2026). The cheapest model for your job depends on how much it writes as well as reads, so sort by output price too, or price a real workload with your own token counts. Ready-made views: cheapest overall, cheapest with a 1M-token context and cheapest with prompt caching.

Why do input and output tokens cost different amounts?

Prices are list prices in US dollars per million tokens. Input covers everything you send (system prompt, history, documents); output is the model’s reply and costs more because it’s generated one token at a time. Cached input is the discounted price for re-reading a prompt prefix the provider has already cached, which matters for chatbots and agents that resend the same instructions on every request. Across the table, output costs a median of 4 times as much as input. A dash means the provider hasn’t published that price.

Which LLM has the largest context window?

Grok 4.20 from xAI, at 2,000,000 tokens (as of 22 November 2026), and 121 models in the table accept a million tokens or more. Paste a document into the context window checker to check which models fit your text.

What do max output and “Fill context” mean?

The context window is the most a single request can hold, prompt and reply together. Max output limits the reply on its own. The Fill context column multiplies the window by the input price, which tells you what one maximum-length prompt costs. Two models can share a 1M-token window and still differ by more than a hundred times on this column, so it’s a quick way to see which long-context models are practical for everyday use.

Is the cheapest model the best choice?

Not on its own. Use the capability filters to shortlist models that meet your hard requirements, then test the shortlist on your own prompts. A cheaper model that needs two attempts, or writes twice as many tokens, can cost more than a dearer one that gets it right first time.

Capabilities and open weights

The capability filters (image input, tool calling, reasoning, prompt caching, structured output) use the features each provider declares for its API. They tell you a feature exists, not how well it works. Open-weight models publish their weights, so several companies host them and you can also run them yourself; their price here is a typical hosted price, and our VRAM calculator covers self-hosting.

Where the data comes from

The table is generated from the OpenRouter models API and the open-source LiteLLM price list, refreshed every day. We correct the data by hand when a source is wrong; for example, when a feed reports a cache price as the normal input price, the model is held back until it’s checked against the provider’s pricing page. Release dates are when a model was first listed by our source, which can differ from the announcement date. Every price move is recorded in the AI pricing changes log, and the methodology page has the details. For one model’s prices, limits and cheaper alternatives, open the AI model pricing pages; to weigh two models against each other, see the head-to-head model comparisons.

Examples

Which AI models are cheapest, and which have the largest context windows?

Largest context windows

Most tokens a single request can hold.

  1. 1Grok 4.20xAI2,000,000 tokens
  2. 2Grok 4.20 Multi-AgentxAI2,000,000 tokens
  3. 3GPT-5.4OpenAI1,050,000 tokens
  4. 4GPT-5.4 ProOpenAI1,050,000 tokens
  5. 5GPT-5.5OpenAI1,050,000 tokens

Cheapest with a 1M-token context

Lowest input price among models that accept at least 1,000,000 tokens.

  1. 1Qwen3.7 FlashQwen$0.03 in · $0.13 out
  2. 2Hy4 flashTencent$0.035 in · $0.1405 out
  3. 3Ling 3.1 FlashinclusionAI$0.06 in · $0.18 out
  4. 4Qwen3.5-FlashQwen$0.065 in · $0.26 out
  5. 5Particle-1Particle AI$0.07 in · $0.50 out

Cheapest with image input and tool calling

For agents that need to see screenshots and call functions.

  1. 1Qwen3.7 FlashQwen$0.03 in · $0.13 out
  2. 2Hy4 flashTencent$0.035 in · $0.1405 out
  3. 3Gemma 3 4BGoogle$0.04 in · $0.05 out
  4. 4Gemma 3 12BGoogle$0.05 in · $0.15 out
  5. 5GPT-5 NanoOpenAI$0.05 in · $0.40 out

Newest releases

Date each model was first listed by our data source.

  1. 1DaisyRose 27BRadixArk2026-11-21
  2. 2Inkling 2 MiniThinking Machines2026-11-21
  3. 3Pixtral Large 2502Mistral2026-11-20
  4. 4Gemini 4 NeonGoogle2026-11-18
  5. 5GPT-6.2 AstraOpenAI2026-11-17

Lists computed from the data on 2026-11-22. Sources: OpenRouter models API and LiteLLM model prices and context windows.

Weighing two models? Head-to-head pages compare them on specs and real workload costs: Claude Sonnet 5.5 vs GPT-6.1 Sol, Claude Opus 5.5 vs Claude Sonnet 5.5, Claude Fable 5.1 vs Claude Opus 5.5, Gemini 3.8 Flash vs GPT-6 Luna, Gemini 3.8 Flash vs GPT-6.1 Sol, Claude Sonnet 5.5 vs Gemini 3.8 Flash and all 22 comparisons.

FAQ

Frequently asked questions

Who has the cheapest AI model?

inclusionAI, on input price: Ling 3.0 Tiny costs $0.015 per million input tokens and $0.045 per million output tokens (as of 22 November 2026). Output price, how many tokens a model turns your text into and whether it is good enough for the task matter as much, so read the cheapest LLM APIs guide and test a shortlist.

Which model has the largest context window?

Grok 4.20, with 2,000,000 tokens (as of 22 November 2026). 121 of the 322 models in the table accept at least 1,000,000 tokens. Every token you send is billed, so a bigger window is only worth paying for when you really need to send that much at once.

Where can I compare AI model benchmarks?

Not here: this table compares prices and limits, not quality. For benchmark scores, use an independent leaderboard such as Artificial Analysis, then come back to price your shortlist and check its context window and features.

What does “Fill context” mean?

It is the price of one request whose prompt uses the model's entire context window, at the list input price (and the long-context price where one applies). It shows how expensive very long prompts get: across the models in the table it ranges from $0.000328 to $63.00 per request.

What is the difference between context window and max output?

The context window is the total number of tokens a request can hold, counting both your prompt and the reply. Max output is the most the model will generate in one reply. A model with a 1,000,000-token window and 65,536 max output can read a huge document but can only write about 50,000 words back in one go.

Why do input and output have different prices?

Reading your prompt can be done in parallel, but generating a reply happens one token at a time, which costs the provider more compute per token. That's why output is usually several times more expensive than input. To see what a real workload costs, use the LLM cost calculator.

What does “open weights” mean?

The model's weights are published, so you can download it and run it on your own hardware or pick from several hosting companies. The price shown for open-weight models is a typical hosted API price; running them yourself costs whatever your hardware and electricity cost.

How current is this table?

Prices, context windows and capabilities are refreshed automatically every day from the OpenRouter models API and the open-source LiteLLM price list, with manual corrections when a source is wrong. This table was last updated on 2026-11-22. See the methodology.

Is the cheapest model the best choice?

Not necessarily. Price says nothing about quality, speed or reliability for your task. Use the filters to shortlist models that meet your hard requirements (context size, image input, tool calling), then test the shortlist on your own prompts before choosing.

Related