An AI token is the smallest unit of text that a large language model processes. In English, one token averages ~4 characters or three-quarters of a word — so "hamburger" tokenizes to 3 tokens ("ham", "bur", "ger") and "hi" is 1 token. AI APIs from Anthropic (Claude), OpenAI (GPT-4o), and Google (Gemini) bill by the token, charging separately for input (text you send) and output (text generated). This convention has held across every major provider since GPT-3 launched in 2020.

What is the difference between input and output tokens?

Input tokens are everything you send to the model — your prompt, system instructions, conversation history, and any retrieved context. Output tokens are what the model generates in response. As of May 2026, output tokens are 3–5× more expensive than input tokens across every major provider: Claude Sonnet 4 is $3/1M input vs $15/1M output (5×), GPT-4o is $2.50/1M input vs $10/1M output (4×), and Gemini 2.0 Flash is $0.10/1M input vs $0.40/1M output (4×). The reason: generating each output token requires a full forward pass through the model, while input tokens are processed in parallel during prefill.

How many tokens is 1,000 words?

1,000 words of English text is approximately 1,333 tokens, based on the standard ratio of ~1.33 tokens per English word (4 characters per token, ~5.3 characters per word including spaces). A 250-word page is ~333 tokens, a 2,500-word document is ~3,333 tokens, and an 80,000-word novel is ~106,667 tokens. Code, non-English languages, and structured data tokenize differently — Chinese and Japanese often run 1–2 tokens per character.

How many tokens is 1 page of text?

A standard single-spaced page of English text (250–300 words) contains ~330–400 tokens. A double-spaced manuscript page (~250 words) is also ~333 tokens. A densely packed US Letter page at 500 words is ~667 tokens. For reference, the entire works of Shakespeare (~835,000 words) come to roughly 1.1 million tokens — fitting comfortably inside a single Gemini 2.5 Pro request (1M context window).

Which AI model is the cheapest in 2026?

As of May 2026, Llama 3.1 8B is the cheapest hosted mainstream model at ~$0.05/1M input tokens, followed by Gemini 1.5 Flash at $0.075/1M, Gemini 2.0 Flash at $0.10/1M, Mistral Small at $0.10/1M, GPT-4o mini at $0.15/1M, and Claude Haiku 4 at $0.25/1M. For high-volume production workloads, Gemini 2.0 Flash usually wins on cost-per-quality because it pairs the $0.10/1M input price with a 1 million-token context window — far larger than the others at the same tier.

What does "per million tokens" mean?

AI APIs price by the million tokens (often written $/1M tokens or $/MTok). If a model lists "$3/1M input", a 1,000-token prompt costs $0.003 (1,000 ÷ 1,000,000 × $3). Most single API calls cost fractions of a cent — but at production scale (10M+ requests/month) the choice between $0.10/1M and $3/1M is the difference between a $30 bill and a $900 bill for the same workload.

How does the TokenRate calculator work?

Enter a budget (USD), a token count, or a character count into the calculator. TokenRate uses live pricing from the OpenRouter API — refreshed hourly via Next.js ISR — to compute the exact tokens-per-dollar or dollars-per-token for every tracked model in parallel. It uses the standard 4-characters-per-token approximation for English text; precise per-tokenizer counts require provider-specific libraries (tiktoken for OpenAI, the Anthropic SDK for Claude).

Is the pricing data real-time?

Live pricing is pulled from the OpenRouter API and refreshed every hour via Next.js Incremental Static Regeneration. OpenRouter aggregates real-time pricing from Anthropic, OpenAI, Google, Meta, DeepSeek, xAI, and Mistral. Reference pricing on individual model pages is reviewed on a rolling cadence (last full sweep: 2026-05-22). For mission-critical decisions, always confirm with the provider's own pricing page.

Can I reduce my AI API bill without changing model?

Yes. Three high-leverage moves: (1) Prompt caching cuts the cost of cached input tokens by ~90% on Anthropic and ~50% on OpenAI — ideal for long, static system prompts. (2) Batch APIs from Anthropic and OpenAI offer a 50% discount with a 24-hour turnaround for asynchronous workloads. (3) Output-length controls (setting max_tokens, requesting concise answers) directly reduce the expensive side of the bill. Combined, these can cut costs 5–10× without changing models.

NodeaBranching AI chat

A branching
AI chat canvas.

How should we launch?

A cautious plan — or a bolder one?

↳ branch the reply

Open a canvas — free

Free during beta · No credit card · nodea.ai

NodeaBranching AI chat

Why settle for
one answer?

Ask once. Branch the reply and explore every option in parallel — then keep the best.

Free during beta · No credit card · nodea.ai

TokenRate · tokenrate.dev

AI Token Calculator & Live Pricing Comparison

TYPE ANY AMOUNT — RESULTS UPDATE LIVE

204 models · 11 providers

Quick

Model	Tier	Quality	Rate in / out / 1M	Tokens In	Tokens Out	Value
Claude Fable 5	Flagship	95	$10.00 – $50.00	—	—	9.5
o3	Reasoning	82	$2.00 – $8.00	—	—	41
Gemini 2.5 Pro	Flagship	76	$1.25 – $10.00	—	—	61
Llama 4 Maverick	Balanced	62	$0.150 – $0.600	—	—	413
R1 Distill Qwen 32B	Reasoning	73	$0.290 – $0.290	—	—	252
Mistral Large	Flagship	50	$2.00 – $6.00	—	—	25
Grok 4.3	Flagship	78	$1.25 – $2.50	—	—	62
Qwen3.7 Max	Flagship	72elo	$1.25 – $3.75	—	—	58
Command A	Balanced	—	$2.50 – $10.00	—	—	—
Nova Pro 1.0	Balanced	—	$0.800 – $3.20	—	—	—
Phi 4	Balanced	—	$0.065 – $0.140	—	—	—
Claude Opus 4.8	Flagship	90	$5.00 – $25.00	—	—	18
GPT-5.1-Codex-Max	Flagship	78	$1.25 – $10.00	—	—	62
Gemini 3.5 Flash	Fast	73elo	$1.50 – $9.00	—	—	49
Llama 4 Scout	Balanced	55	$0.100 – $0.300	—	—	550
R1	Reasoning	73	$0.700 – $2.50	—	—	104
Mistral Medium 3.1	Balanced	48	$0.400 – $2.00	—	—	120
Grok 4.20	Flagship	78	$1.25 – $2.50	—	—	62
Qwen2.5 Coder 32B Instruct	Balanced	58	$0.660 – $1.00	—	—	88
Command R7B (12-2024)	Balanced	—	$0.037 – $0.150	—	—	—
Nova Premier 1.0	Balanced	—	$2.50 – $12.50	—	—	—
WizardLM-2 8x22B	Balanced	—	$0.620 – $0.620	—	—	—
Claude Opus 4.7	Flagship	88	$5.00 – $25.00	—	—	18
GPT-5.1	Flagship	78	$1.25 – $10.00	—	—	62
Nano Banana (Gemini 2.5 Flash Image)	Fast	66	$0.300 – $2.50	—	—	220
Llama 3.3 70B Instruct	Balanced	52	$0.100 – $0.320	—	—	520
R1 Distill Llama 70B	Reasoning	73	$0.800 – $0.800	—	—	91
Mistral Medium 3	Balanced	48	$0.400 – $2.00	—	—	120
Grok 4.20 Multi-Agent	Flagship	78	$2.00 – $6.00	—	—	39
Qwen2.5 72B Instruct	Balanced	53	$0.360 – $0.400	—	—	147
Command R (08-2024)	Balanced	—	$0.150 – $0.600	—	—	—
Nova Micro 1.0	Fast	—	$0.035 – $0.140	—	—	—
Phi 4 Mini Instruct	Fast	—	$0.080 – $0.350	—	—	—
Claude Opus 4.6	Flagship	86	$5.00 – $25.00	—	—	17
GPT-5.1 Chat	Flagship	78	$1.25 – $10.00	—	—	62
Gemini 2.5 Flash	Fast	66	$0.300 – $2.50	—	—	220
Llama 3.1 70B Instruct	Balanced	47	$0.400 – $0.400	—	—	118
DeepSeek V3.1	Balanced	65	$0.210 – $0.790	—	—	310
Mistral Medium 3.5	Balanced	48	$1.50 – $7.50	—	—	32
Grok Build 0.1	Balanced	50	$1.00 – $2.00	—	—	50
Qwen2.5 7B Instruct	Balanced	36	$0.040 – $0.100	—	—	900
Command R+ (08-2024)	Balanced	—	$2.50 – $10.00	—	—	—
Nova Lite 1.0	Fast	—	$0.060 – $0.240	—	—	—
Claude Opus 4.5	Flagship	84	$5.00 – $25.00	—	—	17
GPT-5.1-Codex	Flagship	78	$1.25 – $10.00	—	—	62
Gemini 2.5 Flash Lite	Fast	55	$0.100 – $0.400	—	—	550
Llama 3 70B Instruct	Balanced	44	$0.510 – $0.740	—	—	86
DeepSeek V3.2	Balanced	65	$0.229 – $0.343	—	—	284
Mistral Small 3.2 24B	Balanced	42	$0.075 – $0.200	—	—	560
Qwen3 Max	Flagship	—	$0.780 – $3.90	—	—	—
Nova 2 Lite	Fast	—	$0.300 – $2.50	—	—	—
Claude Opus 4.1	Flagship	80	$15.00 – $75.00	—	—	5.3
GPT-5 Codex	Flagship	78	$1.25 – $10.00	—	—	62
Gemma 3 4B	Balanced	—	$0.050 – $0.100	—	—	—
Llama 3.2 11B Vision Instruct	Balanced	32	$0.345 – $0.345	—	—	93
DeepSeek V3.1 Terminus	Balanced	65	$0.270 – $0.950	—	—	241
Mistral Small 3.1 24B	Balanced	42	$0.351 – $0.555	—	—	120
Qwen3 8B	Balanced	—	$0.050 – $0.400	—	—	—
Claude Sonnet 4.6	Balanced	79	$3.00 – $15.00	—	—	26
GPT-5 Chat	Flagship	78	$1.25 – $10.00	—	—	62
Gemma 3 12B	Balanced	—	$0.050 – $0.150	—	—	—
Llama 3.1 8B Instruct	Balanced	30	$0.020 – $0.030	—	—	1.5k
DeepSeek V3	Balanced	60	$0.200 – $0.800	—	—	300
Mixtral 8x22B Instruct	Balanced	40	$2.00 – $6.00	—	—	20
Qwen3 Coder 30B A3B Instruct	Balanced	—	$0.070 – $0.270	—	—	—
Claude Sonnet 4.5	Balanced	76	$3.00 – $15.00	—	—	25
GPT-5	Flagship	78	$1.25 – $10.00	—	—	62
Gemma 4 26B A4B	Balanced	—	$0.060 – $0.330	—	—	—
Llama 3 8B Instruct	Balanced	26	$0.140 – $0.140	—	—	186
DeepSeek V4 Pro	Balanced	—	$0.435 – $0.870	—	—	—
Mistral Nemo	Balanced	38	$0.020 – $0.030	—	—	1.9k
Qwen3 VL 8B Instruct	Balanced	—	$0.080 – $0.500	—	—	—
Claude Sonnet 4	Balanced	72	$3.00 – $15.00	—	—	24
GPT-5.3 Chat	Flagship	78	$1.75 – $14.00	—	—	45
Gemma 3n 4B	Balanced	—	$0.060 – $0.120	—	—	—
Llama 3.2 3B Instruct	Balanced	22	$0.051 – $0.335	—	—	432
DeepSeek V4 Flash	Fast	—	$0.098 – $0.197	—	—	—
Saba	Balanced	—	$0.200 – $0.600	—	—	—
Qwen3 32B	Balanced	—	$0.080 – $0.280	—	—	—
Claude Opus 4	Flagship	68	$15.00 – $75.00	—	—	4.5
GPT-5.3-Codex	Flagship	78	$1.75 – $14.00	—	—	45
Gemma 3 27B	Balanced	—	$0.080 – $0.160	—	—	—
Llama 3.2 1B Instruct	Balanced	15	$0.027 – $0.201	—	—	556
R1 0528	Reasoning	73	$0.500 – $2.15	—	—	146
Mistral Large 3 2512	Flagship	50	$0.500 – $1.50	—	—	100
Qwen3 Next 80B A3B Instruct	Balanced	—	$0.090 – $1.10	—	—	—
Claude Haiku 4.5	Fast	66	$1.00 – $5.00	—	—	66
GPT-5.2-Codex	Flagship	78	$1.75 – $14.00	—	—	45
Gemma 4 31B	Balanced	—	$0.120 – $0.350	—	—	—
Llama Guard 4 12B	Balanced	—	$0.180 – $0.180	—	—	—
DeepSeek V3 0324	Balanced	65	$0.200 – $0.770	—	—	325
Mistral Large 2407	Flagship	50	$2.00 – $6.00	—	—	25
Qwen3.5-9B	Balanced	—	$0.100 – $0.150	—	—	—
Claude 3.5 Haiku	Fast	55	$0.800 – $4.00	—	—	69
GPT-5.2 Chat	Flagship	78	$1.75 – $14.00	—	—	45
Gemma 2 27B	Balanced	—	$0.650 – $0.650	—	—	—
Llama Guard 3 8B	Balanced	—	$0.484 – $0.030	—	—	—
DeepSeek V3.2 Exp	Balanced	65	$0.270 – $0.410	—	—	241
Codestral 2508	Balanced	50	$0.300 – $0.900	—	—	167
Qwen3 14B	Balanced	—	$0.100 – $0.240	—	—	—
Claude 3 Haiku	Fast	30	$0.250 – $1.25	—	—	120
GPT-5.2	Flagship	78	$1.75 – $14.00	—	—	45
Gemini 3.1 Flash Lite	Fast	—	$0.250 – $1.50	—	—	—
Mistral Small 3	Balanced	42	$0.050 – $0.080	—	—	840
Qwen3 VL 32B Instruct	Balanced	—	$0.104 – $0.416	—	—	—
Claude Opus 4.8 (Fast)	Flagship	90	$10.00 – $50.00	—	—	9.0
GPT-5 Image	Flagship	78	$10.00 – $10.00	—	—	7.8
Gemini 2.5 Pro Preview 06-05	Flagship	76	$1.25 – $10.00	—	—	61
Mistral Small 4	Balanced	42	$0.150 – $0.600	—	—	280
Qwen3 Coder Next	Balanced	—	$0.110 – $0.800	—	—	—
Claude Opus 4.7 (Fast)	Flagship	88	$30.00 – $150.00	—	—	2.9
GPT-5 Pro	Flagship	78	$15.00 – $120.00	—	—	5.2
Gemini 2.5 Pro Preview 05-06	Flagship	76	$1.25 – $10.00	—	—	61
Ministral 3 8B 2512	Balanced	35	$0.150 – $0.150	—	—	233
Qwen3 30B A3B	Balanced	—	$0.120 – $0.500	—	—	—
Claude Opus 4.6 (Fast)	Flagship	86	$30.00 – $150.00	—	—	2.9
GPT-5.2 Pro	Flagship	78	$21.00 – $168.00	—	—	3.7
Gemini 3.1 Pro Preview Custom Tools	Flagship	75elo	$2.00 – $12.00	—	—	38
Ministral 3 3B 2512	Balanced	28	$0.100 – $0.100	—	—	280
Qwen3 VL 30B A3B Instruct	Balanced	—	$0.130 – $0.520	—	—	—
o1	Reasoning	78	$15.00 – $60.00	—	—	5.2
Gemini 3.1 Pro Preview	Flagship	75elo	$2.00 – $12.00	—	—	38
Voxtral Small 24B 2507	Balanced	—	$0.100 – $0.300	—	—	—
Qwen3.5-35B-A3B	Balanced	—	$0.140 – $1.00	—	—	—
GPT-5.1-Codex-Mini	Fast	78	$0.250 – $2.00	—	—	312
Nano Banana Pro (Gemini 3 Pro Image Preview)	Fast	75elo	$2.00 – $12.00	—	—	38
Ministral 3 14B 2512	Balanced	—	$0.200 – $0.200	—	—	—
Qwen3.6 35B A3B	Balanced	—	$0.150 – $1.00	—	—	—
GPT-5 Image Mini	Fast	78	$2.50 – $2.00	—	—	31
Gemini 3 Flash Preview	Fast	72elo	$0.500 – $3.00	—	—	144
Devstral 2 2512	Balanced	—	$0.400 – $2.00	—	—	—
Qwen3.5-27B	Balanced	—	$0.195 – $1.56	—	—	—
o4 Mini High	Reasoning	75	$1.10 – $4.40	—	—	68
Gemini 2.5 Flash Lite Preview 09-2025	Fast	55	$0.100 – $0.400	—	—	550
Qwen3 VL 235B A22B Instruct	Balanced	—	$0.200 – $0.880	—	—	—
o4 Mini	Reasoning	75	$1.10 – $4.40	—	—	68
Gemini 3.1 Flash Lite Preview	Fast	—	$0.250 – $1.50	—	—	—
Qwen3 Coder 480B A35B	Balanced	—	$0.220 – $1.80	—	—	—
o4 Mini Deep Research	Reasoning	75	$2.00 – $8.00	—	—	38
Nano Banana 2 (Gemini 3.1 Flash Image Preview)	Fast	—	$0.500 – $3.00	—	—	—
Qwen2.5 VL 72B Instruct	Balanced	—	$0.250 – $0.750	—	—	—
GPT-5.4	Flagship	73elo	$2.50 – $15.00	—	—	29
Qwen3.5-122B-A10B	Balanced	—	$0.260 – $2.08	—	—	—
GPT-5.4 Image 2	Flagship	73elo	$8.00 – $15.00	—	—	9.1
Qwen3.5 Plus 2026-02-15	Balanced	—	$0.260 – $1.56	—	—	—
GPT-5.4 Pro	Flagship	73elo	$30.00 – $180.00	—	—	2.4
Qwen-Plus	Balanced	—	$0.260 – $0.780	—	—	—
GPT-5.4 Nano	Fast	73elo	$0.200 – $1.25	—	—	365
Qwen3.6 27B	Balanced	—	$0.289 – $2.40	—	—	—
GPT-5.4 Mini	Fast	73elo	$0.750 – $4.50	—	—	97
Qwen3.5 Plus 2026-04-20	Balanced	—	$0.300 – $1.80	—	—	—
GPT-5.5	Flagship	72elo	$5.00 – $30.00	—	—	14
Qwen3.7 Plus	Balanced	—	$0.320 – $1.28	—	—	—
GPT-5.5 Pro	Flagship	72elo	$30.00 – $180.00	—	—	2.4
Qwen3.6 Plus	Balanced	—	$0.325 – $1.95	—	—	—
o3 Mini High	Reasoning	72	$1.10 – $4.40	—	—	65
Qwen3.5 397B A17B	Balanced	—	$0.390 – $2.34	—	—	—
o3 Mini	Reasoning	72	$1.10 – $4.40	—	—	65
Qwen3 235B A22B	Balanced	—	$0.455 – $1.82	—	—	—
GPT-5 Mini	Fast	68	$0.250 – $2.00	—	—	272
Qwen3 Coder Plus	Balanced	—	$0.650 – $3.25	—	—	—
GPT-4o	Balanced	64	$2.50 – $10.00	—	—	26
Qwen3.5-Flash	Fast	—	$0.065 – $0.260	—	—	—
GPT-4.1	Flagship	62	$2.00 – $8.00	—	—	31
Qwen3.6 Flash	Fast	—	$0.188 – $1.13	—	—	—
GPT-4.1 Nano	Fast	62	$0.100 – $0.400	—	—	620
Qwen3 Coder Flash	Fast	—	$0.195 – $0.975	—	—	—
GPT-4 Turbo	Balanced	58	$10.00 – $30.00	—	—	5.8
Qwen3.6 Max Preview	Flagship	—	$1.04 – $6.24	—	—	—
GPT-4	Balanced	58	$30.00 – $60.00	—	—	1.9
Qwen3 30B A3B Thinking 2507	Reasoning	—	$0.080 – $0.400	—	—	—
GPT-5 Nano	Fast	55	$0.050 – $0.400	—	—	1.1k
Qwen3 Next 80B A3B Thinking	Reasoning	—	$0.098 – $0.780	—	—	—
GPT-4.1 Mini	Fast	52	$0.400 – $1.60	—	—	130
Qwen3 235B A22B Thinking 2507	Reasoning	—	$0.100 – $0.100	—	—	—
GPT-4o-mini	Fast	51	$0.150 – $0.600	—	—	340
Qwen3 VL 8B Thinking	Reasoning	—	$0.117 – $1.36	—	—	—
GPT-3.5 Turbo	Balanced	35	$0.500 – $1.50	—	—	70
Qwen3 VL 30B A3B Thinking	Reasoning	—	$0.130 – $1.56	—	—	—
GPT-3.5 Turbo Instruct	Balanced	35	$1.50 – $2.00	—	—	23
Qwen3 VL 235B A22B Thinking	Reasoning	—	$0.260 – $2.60	—	—	—
GPT-3.5 Turbo 16k	Balanced	35	$3.00 – $4.00	—	—	12
Qwen Plus 0728 (thinking)	Reasoning	—	$0.260 – $0.780	—	—	—
o3 Deep Research	Reasoning	—	$10.00 – $40.00	—	—	—
Qwen3 Max Thinking	Reasoning	—	$0.780 – $3.90	—	—	—
o3 Pro	Reasoning	—	$20.00 – $80.00	—	—	—
Qwen3 30B A3B Instruct 2507	Balanced	—	$0.048 – $0.193	—	—	—
o1-pro	Reasoning	—	$150.00 – $600.00	—	—	—
Qwen3 235B A22B Instruct 2507	Balanced	—	$0.090 – $0.100	—	—	—
gpt-oss-20b	Balanced	—	$0.029 – $0.140	—	—	—
Qwen Plus 0728	Balanced	—	$0.260 – $0.780	—	—	—
gpt-oss-120b	Balanced	—	$0.039 – $0.180	—	—	—
gpt-oss-safeguard-20b	Balanced	—	$0.075 – $0.300	—	—	—
GPT Audio	Balanced	—	$2.50 – $10.00	—	—	—
GPT Audio Mini	Fast	—	$0.600 – $2.40	—	—	—
GPT-4o Search Preview	Balanced	64	$2.50 – $10.00	—	—	26
GPT-4o (2024-11-20)	Balanced	64	$2.50 – $10.00	—	—	26
GPT-4o (2024-08-06)	Balanced	64	$2.50 – $10.00	—	—	26
GPT-4o (2024-05-13)	Balanced	64	$5.00 – $15.00	—	—	13
GPT-4 Turbo Preview	Balanced	58	$10.00 – $30.00	—	—	5.8
GPT-4o-mini Search Preview	Fast	51	$0.150 – $0.600	—	—	340
GPT-4o-mini (2024-07-18)	Fast	51	$0.150 – $0.600	—	—	340
GPT-3.5 Turbo (older v0613)	Balanced	35	$1.00 – $2.00	—	—	35
GPT Chat Latest	Balanced	—	$5.00 – $30.00	—	—	—

Approximation: ~4 chars / 1 token · Rates shown per 1M tokens · Pricing is live via OpenRouter — confirm with provider before billing decisions.

Quick

204 models

↑ Type any amount — results update live

~4 chars ≈ 1 token · rates per 1M tokens · live via OpenRouter — confirm with provider before billing.

222+Models tracked

LivePricing data

FreeNo sign-up

Free, instant cost estimates for every major AI model. Convert money, tokens, and characters across Claude, GPT-4o, Gemini, Llama, DeepSeek, Grok, and Mistral — with live pricing from OpenRouter.

What is an AI token calculator? An AI token calculator converts between dollars, token counts, and characters using live API pricing from large language model providers. It answers two questions developers ask before shipping with an LLM: "how much will this prompt cost?" and "how many tokens does my budget buy?"

Current AI API pricing (June 2026): Llama 3.1 8B costs $0.02/1M input tokens; Gemini 2.0 Flash costs $0.10/1M; GPT-4o mini costs $0.15/1M; Claude Haiku 4 costs $0.25/1M. Output tokens cost roughly 3–5× the input price across every major provider. A typical 1,000-token request on a balanced model like Claude Sonnet 4 ($3/1M input) costs $0.003 — under a third of a cent. Use the calculator above to convert any amount across 222+ tracked models.

Prices verified 2026-06-11 · Live data refreshed hourly from OpenRouter

What Are AI Tokens?

A token is the basic unit that AI language models use to process text. Rather than reading word by word, models break text into token fragments — roughly 4 characters each for English. The word "hamburger" is 3 tokens; "hi" is 1 token.

Every AI API — including Claude, GPT-4o, and Gemini — charges by the token. You pay for the text you send (input tokens) and the text the model generates (output tokens). Understanding tokens is the key to predicting and controlling your AI costs.

Read the full guide →

How AI Token Pricing Works

AI models are priced by the million tokens — written as $/1M tokens. The two numbers you always see are:

InputThe text you send — your prompt, system instructions, and conversation history.
OutputThe text the model generates. Usually 3–5× more expensive than input, because generating tokens requires more compute than reading them.

A typical 1,000-token request (about 750 words of prompt) with Claude Sonnet 4 costs $0.003 — less than a third of a cent. At scale, though, those fractions add up fast.

Read the full pricing guide →

Cheapest AI Models in 2026

Ranked by input token price. All prices per 1 million tokens.

Model	Provider	Input / 1M	Output / 1M	Context
Llama 3.1 8B	Meta	$0.020	$0.030	131K
Mistral Nemo	Mistral	$0.020	$0.030	131K
Llama 3.2 1B	Meta	$0.027	$0.201	131K
gpt-oss-20b	OpenAI	$0.029	$0.140	131K
Amazon Nova Micro	Amazon	$0.035	$0.140	128K
Command R7B (12-2024)	Cohere	$0.037	$0.150	128K

Prices are approximate and may change. See full comparison →

Token Calculation Examples

1,000 words of English text

~1,333 tokens

~4 chars per token

A standard email (150 words)

~200 tokens

under $0.001 on most models

A 10-page PDF (2,500 words)

~3,333 tokens

$0.01 with Claude Sonnet 4

An entire novel (80,000 words)

~107K tokens

fits in most flagship models

A full codebase (100K lines)

~300K+ tokens

needs 200K+ context window

1 million tokens budget at $1

~750K words

with Gemini 2.0 Flash

Want more? See the full token reference guide →

Browse All AI Models

Detailed pricing pages for every major model — with context windows, strengths, use cases, and cost examples.

Claude Opus 4Anthropic$15.00/1M in Claude Sonnet 4Anthropic$3.00/1M in Claude Haiku 4Anthropic$0.250/1M in Claude 3.5 SonnetAnthropic$3.00/1M in Claude 3 HaikuAnthropic$0.250/1M in Claude 3.7 SonnetAnthropic$3.00/1M in Claude 3.5 HaikuAnthropic$0.800/1M in GPT-4oOpenAI$2.50/1M in GPT-4o miniOpenAI$0.150/1M in

Frequently Asked Questions

Related Tools & Guides

Words → Tokens ConverterPaste any text and see the token count instantly.Token → USD CalculatorConvert a raw token count to dollars for any model.Cheapest AI Models 2026Ranked list of the most affordable LLMs available via API.What Are AI Tokens?A clear explanation of how tokens work and why they matter.Quality Per Dollar: Best-Value LLMsWhich models deliver the most quality for each dollar in 2026.Gemini vs Claude vs GPT CostA full side-by-side cost comparison of the three frontier model families.LLM Pricing Trends 2026How AI model costs have shifted across providers this year.

A branchingAI chat canvas.

Why settle forone answer?