Claude vs ChatGPT API: Which One Cheap Pass?

Di short answer

Claude vs ChatGPT API pricing no get one clear winner, because both vendors dey charge input and output differently for every tier. As of August 2026, the top two Claude tiers dey charge the same amount per input token as their OpenAI counterparts, but dem dey charge less per output token. So Claude dey slightly cheaper for work wey output plenty. For the small tier, Claude cost several times more for the same job. Any answer wey no mention tier and token mix na guess.

So this page go first show you the calculation, then three workloads with their token mixes and costs, then the multipliers wey fit increase real bill pass the headline rate.

Formula wey you need

Both APIs dey bill text the same way. You pay for tokens wey you send, and you pay higher rate for tokens wey the model write back.

cost = (input tokens / 1,000,000) * input rate + (output tokens / 1,000,000) * output rate

One token na about 4 characters for English, close to 0.75 of one word. Use this only for first estimate. For real bill, use the counts wey each API return inside the usage object for every response.

# Rates are US dollars per million tokens.
def cost(in_tok, out_tok, in_rate, out_rate):
    return in_tok / 1e6 * in_rate + out_tok / 1e6 * out_rate

# One support-chat turn: 1,400 tokens in, 220 tokens out, on a $2 / $10 model.
print(round(cost(1400, 220, 2.0, 10.0), 6))

The script dey print 0.005, wey be half cent for each turn. Multiply am by the number of turns wey you expect for one month, and you get budget. If the million-token unit for that denominator still dey look abstract, wetin one million tokens fit buy for Claude go show you concrete amount of text behind am. Learn this one formula, and every price change after today go be one-line edit instead of new analysis.

Published rates, August 2026

Na be the list prices wey both vendors publish on 1 August 2026. This block na the only place for this post wey price dey written down. Check am against claude.com/pricing and OpenAI pricing page before you commit budget.

[
  {
    "label": "Frontier",
    "claude_input": 5,
    "claude_output": 25,
    "openai_input": 5,
    "openai_output": 30
  },
  {
    "label": "Workhorse",
    "claude_input": 2,
    "claude_output": 10,
    "openai_input": 2,
    "openai_output": 12
  },
  {
    "label": "Workhorse from September",
    "claude_input": 3,
    "claude_output": 15,
    "openai_input": 2,
    "openai_output": 12
  },
  {
    "label": "Small",
    "claude_input": 1,
    "claude_output": 5,
    "openai_input": 0.2,
    "openai_output": 1.2
  }
]

The pairings follow position for each lineup, no be fame. Frontier na Claude Opus 5 against gpt-5.6-sol. Workhorse na Claude Sonnet 5 against gpt-5.6-terra. Small na Claude Haiku 4.5 against gpt-5.6-luna. Both vendors still sell higher tiers above this block, and OpenAI still list older generations with their original rates. For Claude, that tier na Fable 5. E cost double the frontier row for both legs, while e keep the same five-times output ratio. So wetin Fable 5 cost and when e deserve that rate worth reading before you assume say Opus 5 na the highest item for your bill. If you compare flagship for one side with budget model for the other, the number no mean anything. Na so most published comparisons get their headline.

Two things for that block need clear explanation. Sonnet 5's 2 and 10 na introductory rates. Anthropic document say dem end on 31 August 2026. After that, Sonnet 5 go charge 3 and 15. That one change turn the workhorse tier from small Claude win to clear OpenAI win, and e go happen within the time wey this post remain useful.

The second thing na pattern wey you suppose remember: Claude price output at exactly 5 times its own input rate for every row above, while OpenAI price output at 6 times. Na that ratio, not the absolute rate, dey make the winner change based on your token mix.

Why rate per million tokens no fit compare across vendors

Rate na dollars per token. Bill na dollars per token multiplied by tokens, and token count for one piece of text na property of the model, no be property of the text.

Anthropic document this directly: Claude 4.7 and later models, wey include Claude Opus 5 and Claude Sonnet 5, dey use newer tokenizer wey dey produce roughly 30% more tokens for the same text than Claude Sonnet 4.6 and earlier. The rate no change. The bill change.

So two rates wey look the same no be two identical bills. Count tokens for both sides, using your own text, before you trust any comparison, including this one.

curl https://api.anthropic.com/v1/messages/count_tokens \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-5",
    "messages": [{"role": "user", "content": "Summarise the attached contract."}]
  }'

Healthy response na small JSON object wey hold input_tokens. A 401 mean say the key variable empty for your shell. Run the same text through the other vendor tokenizer, divide one count by the other, then multiply that ratio by the published rate before you compare. If the ratio favour one vendor by more than about 1.2, e outweigh every rate difference for the block above.

Scenario one: chat feature

A support assistant wey dey inside web application. E dey handle 120,000 turns every month. Each turn dey send system prompt, tool schemas, two retrieved help-desk snippets, and some earlier messages. Na about 1,400 input tokens be that, and the model dey answer with about 220 tokens. E dey interactive, so batch discount no fit apply here. Run am for workhorse tier.

[
  {
    "label": "List price",
    "claude_usd": "600.00",
    "openai_usd": "652.80"
  },
  {
    "label": "Cached prefix",
    "claude_usd": "427.20",
    "openai_usd": "480.00"
  },
  {
    "label": "Small tier, cached",
    "claude_usd": "213.60",
    "openai_usd": "48.00"
  }
]

For list price, Claude dey bill 600.00, while OpenAI dey bill 652.80. The input costs dey the same because both input rates dey the same. The full difference come from output, and Claude win am with margin wey too small to use choose vendor.

Now mark 800 of the 1,400 input tokens as stable prefix: the system prompt and tool schemas wey no dey change between turns. Both vendors dey charge 10% of their base input rate when cache hit happen. Claude bill go drop to 427.20, and OpenAI own go drop to 480.00. Caching dey save more money pass vendor choice, and both platforms support am.

Na here Claude lose. Most support answers no need workhorse model. Move the same traffic go small tier, and Claude go bill 213.60, while OpenAI go bill 48.00. Divide one by the other, and Claude cost about four and a half times more for the same work. Claude Haiku 4.5 dey charge 1 per million input tokens, while gpt-5.6-luna dey charge 0.2. No amount of caching fit close fivefold gap. If your workload fit run on small model, OpenAI cheaper, and the margin dey large.

Scenario two: document pipeline wey dey carry long context

25,000 contracts for one month. Every request carry the same 9,000-token instruction and JSON schema, then 12,000 tokens of contract text, and the model return about 700 tokens of structured fields. Output na less than 4% of all the tokens for this job, and na this one fact decide the comparison.

[
  {
    "label": "Chat feature",
    "claude_usd": "427.20",
    "openai_usd": "480.00"
  },
  {
    "label": "Document pipeline",
    "claude_usd": "820.00",
    "openai_usd": "855.00"
  },
  {
    "label": "Coding agent",
    "claude_usd": "116.10",
    "openai_usd": "124.20"
  }
]

With the 9,000-token prefix cached for both sides, Claude charge 820.00 and OpenAI charge 855.00. The input costs match to the cent, because both workhorse input rates match for August 2026. The whole difference come from the output cost, wey Claude price at 5 times input and OpenAI price at 6 times input.

This job fit use batch too. Both vendors give 50% discount on input and output for asynchronous work, so both totals go reduce by half and the gap keep the same shape. Batch na the biggest single discount for either platform, and nightly document pipeline fit always use am.

Claude lose for the same tier logic wey we see before. Field extraction against fixed schema na exactly the kind work wey small model dey handle well, and input na 97% of the tokens here. Moving down one tier multiply OpenAI input rate by 0.1 and Claude input rate by 0.5. Test the small model with 200 documents before you assume say you need the bigger one, using the same evaluation wey you go run for choosing the correct Claude tier for a job.

Scenario three: coding agent wey dey always-on

One developer dey run agent for VPS, roughly 900 model turns every month. Each turn carry about 60,000 input tokens from repository context, and 80% of dem na cache hit. E dey produce about 1,800 tokens of edits and explanation. Price am for frontier tier, because na for coding capability gap fit cost real money.

Claude billing na 116.10, compared with OpenAI wey be 124.20. Input rate dey the same, cache read rate dey the same, and Claude lower output rate win by about 7%.

That 7% dey inside tokenizer error bar wey we talk earlier, so treat this scenario as tie for rates. But billing model no be tie. For one developer at this volume, subscription seat normally cost less than metered API calls, and the whole comparison change once you include subscription. Work through the comparison between API and subscription cost before you put agent on metered billing. If you use metered billing, first learn where the tokens dey go, because coding agent token usage mostly come from context wey e resend for every turn, no be from the code wey e write.

The multipliers wey dey decide the bill

The headline rate na the smallest item for this list. Each one wey follow fit change real bill pass the difference between the two vendors for the same tier.

Cached input. Both vendors charge 10% of base input for cache hit. Anthropic still charge to write the cache: 1.25 times base input for five-minute entry and 2 times for one-hour entry. Later hit go refresh that entry at the read price. So, five-minute entry don pay for itself after one read. Workload wey get long gaps between requests dey write more often than e dey read. For that kind workload, caching fit cost more than wetin e save. Steady traffic dey work well with cache. Bursty traffic no dey.

Batch discounts. Na 50% discount on input and output for both platforms, but na asynchronous work only. If the job fit wait, this one go cut the bill for either vendor by half, and e still fit combine with caching.

Output share. Claude charge 5 times its input rate for output. OpenAI charge 6 times. The more tokens wey your system write instead of read, the better Claude go look when input rate equal. For work wey input dominate, na the reverse go happen.

Retries. Retry dey pay for the whole input again, but e return nothing wey you fit use. If 6% of your calls fail schema validation and get retried, your input leg don become 6% bigger than wetin your plan talk. Log retries as cost line, no be error line. Teams dey discover this multiplier last, and e often pass the vendor gap wey dem spend one week dey argue about.

Which side lose, and for where

Claude lose the small tier straight. gpt-5.6-luna charge 0.2 per million input tokens, while Claude Haiku 4.5 dey charge 1. For output, dem dey charge 1.2 against 5. High-volume classification and short extraction cost roughly four times more for Claude.

Claude lose the workhorse tier from 1 September 2026, when the introductory rate end and Sonnet 5 move to 3 and 15 against gpt-5.6-terra wey price no change. Claude win the frontier tier for output rate, but the margin fit disappear once you count your own tokens. For work wey input dominate, nobody win for August 2026, because both sides dey charge 2 per million and the input price na the same.

How to measure dis on your own traffic

Read the usage object for every response and store four numbers for each request: input tokens, output tokens, cache read tokens, and cache write tokens. For Claude API, dem dey show as input_tokens, output_tokens, cache_read_input_tokens, and cache_creation_input_tokens. Add dem together for each day, multiply by the current rates, then compare the total with the invoice. If both numbers no match, na usually retries or code path wey you forget say you release.

Run the comparison with one week of real traffic instead of sample prompt, and compare cost per completed task instead of cost per token. Model wey answer for one attempt at twice the rate fit cheaper than model wey need three attempts at half the rate. Cost per token na rate. Cost per completed task na your bill.

Set a hard spend ceiling before you leave anything running. Both platforms get spend limits for their consoles, and agent wey stick for retry loop fit spend one month budget overnight. Connect am the same way you go connect cost control for agent wey dey always on. If you dey build the first version for server wey you own, first Claude API app for VPS explain the key handling and request loop wey this calculation dey depend on.

FAQ

Claude API dey cheaper pass ChatGPT API?

No be every time. For August 2026, frontier and workhorse tiers charge the same amount per input token for both sides, and Claude charge less per output token. So Claude fit cost few percent less for work wey get plenty output. For small tier, OpenAI gpt-5.6-luna cost one fifth of Claude Haiku 4.5 per input token. So high-volume classification dey far cheaper for OpenAI. Calculate your own figure from your own token mix, because na your workload, no be price list, go decide such small difference.

Why same words dey cost different number of tokens for each API?

Because every model get its own tokenizer, and model na wetin determine token count. Anthropic document say Claude 4.7 and later models produce about 30% more tokens for the same text than Claude Sonnet 4.6 and earlier. Send your own text to each vendor token counting endpoint. Divide one count by the other, then multiply that ratio by the published rate before you compare. 20% token difference dey bigger than most published rate differences.

Prompt caching fit ever make bill increase?

Yes, for Anthropic. Cache write cost 1.25 times the base input rate for five-minute entry, and 2 times for one-hour entry. Hit cost 0.1 times. If your traffic dey come in bursts and most entries expire before anything read dem, you go pay the write premium and get no reads. Compare cache_creation_input_tokens with cache_read_input_tokens for your logs. Ratio wey near 1 mean say caching dey cost you money. Make the entry last longer or remove am.

I suppose run coding agent for API or use subscription?

For one developer wey dey use am at ordinary volume, subscription seat usually cost less than metered API billing. This na because coding agent dey resend large context for every turn, and dem charge you for that context every time. Measure the usage for one week first. Then compare the API total with the seat price. API better when usage dey come in bursts, or when you need programmatic access wey seat no provide. If na hitting usage limit, instead of price, push you toward API, ways to escape Claude usage limit include several options wey cost less than moving to metered billing.