Compare LLM models

Head-to-head pricing, context window, and tokenizer for the model pairs people actually choose between.

GPT-5.6 Sol vs Claude Opus 5

The flagship from each provider, and what the difference costs.$5.00 vs $5.00 per 1M input

Attribute	GPT-5.6 Sol	Claude Opus 5
Input, per 1M tokens	$5.00	$5.00
Output, per 1M tokens	$30.00	$25.00
Context window	1,050,000	1,000,000
Provider	OpenAI	Anthropic
Tokenizer	o200k_base	Anthropic tokenizer
Long-context tier	above 272,000	none
Retires	no date announced	no date announced

Reading more than you write: Both $5 per million input. Input rate cannot separate this pair.
Writing more than you read: GPT-5.6 Sol prices output at 6x its input rate, Claude Opus 5 at 5x — so Claude Opus 5's lead narrows as answers lengthen.
Whole documents in one request: Claude Opus 5 stops at 1,000,000; GPT-5.6 Sol takes 1,050,000. Past the lower ceiling only GPT-5.6 Sol accepts the request, at any price.
Past the long-context threshold: The ranking inverts (GPT-5.6 Sol above 272,000): GPT-5.6 Sol wins below the line, Claude Opus 5 above it at $5 against $10. Judging this pair on headline rates alone gives the wrong answer for large prompts.
10,000 in, 1,000 out: $0.0800 on GPT-5.6 Sol, $0.0750 on Claude Opus 5. Over 10,000 such calls that is $50.00 in Claude Opus 5's favour.
How these counts were arrived at: GPT-5.6 Sol is counted exactly in your browser; Claude Opus 5 is estimated from a characters-per-token ratio. Close, not exact.

GPT-5.6 Sol details Claude Opus 5 details

GPT-5.6 Terra vs Claude Sonnet 5

The mid-tier option from each provider, where most production work lands.$2.00 vs $2.00 per 1M input

Attribute	GPT-5.6 Terra	Claude Sonnet 5
Input, per 1M tokens	$2.00	$2.00
Output, per 1M tokens	$12.00	$10.00
Context window	1,050,000	1,000,000
Provider	OpenAI	Anthropic
Tokenizer	o200k_base	Anthropic tokenizer
Long-context tier	above 272,000	none
Retires	no date announced	no date announced

Reading more than you write: Both $2 per million input. Input rate cannot separate this pair.
Writing more than you read: GPT-5.6 Terra prices output at 6x its input rate, Claude Sonnet 5 at 5x — so Claude Sonnet 5's lead narrows as answers lengthen.
Whole documents in one request: Claude Sonnet 5 stops at 1,000,000; GPT-5.6 Terra takes 1,050,000. Past the lower ceiling only GPT-5.6 Terra accepts the request, at any price.
Past the long-context threshold: The ranking inverts (GPT-5.6 Terra above 272,000): GPT-5.6 Terra wins below the line, Claude Sonnet 5 above it at $2 against $4. Judging this pair on headline rates alone gives the wrong answer for large prompts.
10,000 in, 1,000 out: $0.0320 on GPT-5.6 Terra, $0.0300 on Claude Sonnet 5. Over 10,000 such calls that is $20.00 in Claude Sonnet 5's favour.
How these counts were arrived at: GPT-5.6 Terra is counted exactly in your browser; Claude Sonnet 5 is estimated from a characters-per-token ratio. Close, not exact.

GPT-5.6 Terra details Claude Sonnet 5 details

GPT-5.6 Luna vs Claude Haiku 4.5

The cheapest current model from each provider, for high-volume work.$0.20 vs $1.00 per 1M input

Attribute	GPT-5.6 Luna	Claude Haiku 4.5
Input, per 1M tokens	$0.20	$1.00
Output, per 1M tokens	$1.20	$5.00
Context window	1,050,000	200,000
Provider	OpenAI	Anthropic
Tokenizer	o200k_base	Anthropic tokenizer
Long-context tier	above 272,000	none
Retires	no date announced	no date announced

Reading more than you write: GPT-5.6 Luna $0.20 against Claude Haiku 4.5 $1 per million input tokens — a wide gap of 5x.
Writing more than you read: GPT-5.6 Luna prices output at 6x its input rate, Claude Haiku 4.5 at 5x — so GPT-5.6 Luna's lead widens as answers lengthen.
Whole documents in one request: Claude Haiku 4.5 stops at 200,000; GPT-5.6 Luna takes 1,050,000. Past the lower ceiling only GPT-5.6 Luna accepts the request, at any price.
Past the long-context threshold: GPT-5.6 Luna above 272,000 bills the whole request higher — $0.40 and $1. GPT-5.6 Luna stays ahead either side of that line.
10,000 in, 1,000 out: $0.0032 on GPT-5.6 Luna, $0.0150 on Claude Haiku 4.5. Over 10,000 such calls that is $118.00 in GPT-5.6 Luna's favour.
How these counts were arrived at: GPT-5.6 Luna is counted exactly in your browser; Claude Haiku 4.5 is estimated from a characters-per-token ratio. Close, not exact.

GPT-5.6 Luna details Claude Haiku 4.5 details

Claude Opus 5 vs Claude Sonnet 5

Whether the Opus tier earns its price over Sonnet.$5.00 vs $2.00 per 1M input

Attribute	Claude Opus 5	Claude Sonnet 5
Input, per 1M tokens	$5.00	$2.00
Output, per 1M tokens	$25.00	$10.00
Context window	1,000,000	1,000,000
Provider	Anthropic	Anthropic
Tokenizer	Anthropic tokenizer	Anthropic tokenizer
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: Claude Sonnet 5 $2 against Claude Opus 5 $5 per million input tokens — a material gap of 2.5x.
10,000 in, 1,000 out: $0.0750 on Claude Opus 5, $0.0300 on Claude Sonnet 5. Over 10,000 such calls that is $450.00 in Claude Sonnet 5's favour.

Claude Opus 5 details Claude Sonnet 5 details

GPT-5.6 Sol vs GPT-5.6 Terra

How far down the GPT-5.6 range you can go before it costs you.$5.00 vs $2.00 per 1M input

Attribute	GPT-5.6 Sol	GPT-5.6 Terra
Input, per 1M tokens	$5.00	$2.00
Output, per 1M tokens	$30.00	$12.00
Context window	1,050,000	1,050,000
Provider	OpenAI	OpenAI
Tokenizer	o200k_base	o200k_base
Long-context tier	above 272,000	above 272,000
Retires	no date announced	no date announced

Reading more than you write: GPT-5.6 Terra $2 against GPT-5.6 Sol $5 per million input tokens — a material gap of 2.5x.
Past the long-context threshold: GPT-5.6 Sol above 272,000, GPT-5.6 Terra above 272,000 bills the whole request higher — $4 and $10. GPT-5.6 Terra stays ahead either side of that line.
10,000 in, 1,000 out: $0.0800 on GPT-5.6 Sol, $0.0320 on GPT-5.6 Terra. Over 10,000 such calls that is $480.00 in GPT-5.6 Terra's favour.

GPT-5.6 Sol details GPT-5.6 Terra details

GPT-4o vs GPT-4o mini

The classic cost-versus-capability split inside one family.$2.50 vs $0.15 per 1M input

Attribute	GPT-4o	GPT-4o mini
Input, per 1M tokens	$2.50	$0.15
Output, per 1M tokens	$10.00	$0.60
Context window	128,000	128,000
Provider	OpenAI	OpenAI
Tokenizer	o200k_base	o200k_base
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: GPT-4o mini $0.15 against GPT-4o $2.50 per million input tokens — an order-of-magnitude gap of 16.7x.
10,000 in, 1,000 out: $0.0350 on GPT-4o, $0.0021 on GPT-4o mini. Over 10,000 such calls that is $329.00 in GPT-4o mini's favour.

GPT-4o details GPT-4o mini details

GPT-4 Turbo vs GPT-5.6 Sol

What migrating off a retiring model actually changes.$10.00 vs $5.00 per 1M input

Attribute	GPT-4 Turbo	GPT-5.6 Sol
Input, per 1M tokens	$10.00	$5.00
Output, per 1M tokens	$30.00	$30.00
Context window	128,000	1,050,000
Provider	OpenAI	OpenAI
Tokenizer	cl100k_base	o200k_base
Long-context tier	none	above 272,000
Retires	2026-10-23	no date announced

Reading more than you write: GPT-5.6 Sol $5 against GPT-4 Turbo $10 per million input tokens — a material gap of twice.
Whole documents in one request: GPT-4 Turbo stops at 128,000; GPT-5.6 Sol takes 1,050,000. Past the lower ceiling only GPT-5.6 Sol accepts the request, at any price.
Past the long-context threshold: The ranking inverts (GPT-5.6 Sol above 272,000): GPT-5.6 Sol wins below the line, GPT-4 Turbo above it at $10 against $10. Judging this pair on headline rates alone gives the wrong answer for large prompts.
10,000 in, 1,000 out: $0.1300 on GPT-4 Turbo, $0.0800 on GPT-5.6 Sol. Over 10,000 such calls that is $500.00 in GPT-5.6 Sol's favour.
One of these is going away: GPT-4 Turbo retires 23 October 2026; GPT-5.6 Sol has no date. Removed models fail outright rather than degrade, so past that date price is irrelevant.

GPT-4 Turbo details GPT-5.6 Sol details

GPT-4o vs GPT-5.6 Terra

Whether moving to the current generation is cheaper as well as newer.$2.50 vs $2.00 per 1M input

Attribute	GPT-4o	GPT-5.6 Terra
Input, per 1M tokens	$2.50	$2.00
Output, per 1M tokens	$10.00	$12.00
Context window	128,000	1,050,000
Provider	OpenAI	OpenAI
Tokenizer	o200k_base	o200k_base
Long-context tier	none	above 272,000
Retires	no date announced	no date announced

Reading more than you write: GPT-5.6 Terra $2 against GPT-4o $2.50 per million input tokens — a modest gap of 1.3x.
Writing more than you read: This flips the pair: GPT-4o is dearer on input but cheaper on output, $10 against $12. Which one costs less depends entirely on your answer length.
Whole documents in one request: GPT-4o stops at 128,000; GPT-5.6 Terra takes 1,050,000. Past the lower ceiling only GPT-5.6 Terra accepts the request, at any price.
Past the long-context threshold: The ranking inverts (GPT-5.6 Terra above 272,000): GPT-5.6 Terra wins below the line, GPT-4o above it at $2.50 against $4. Judging this pair on headline rates alone gives the wrong answer for large prompts.
10,000 in, 1,000 out: $0.0350 on GPT-4o, $0.0320 on GPT-5.6 Terra. Over 10,000 such calls that is $30.00 in GPT-5.6 Terra's favour.

GPT-4o details GPT-5.6 Terra details

GPT-4o vs Claude Sonnet 5

The comparison most teams are actually choosing between.$2.50 vs $2.00 per 1M input

Attribute	GPT-4o	Claude Sonnet 5
Input, per 1M tokens	$2.50	$2.00
Output, per 1M tokens	$10.00	$10.00
Context window	128,000	1,000,000
Provider	OpenAI	Anthropic
Tokenizer	o200k_base	Anthropic tokenizer
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: Claude Sonnet 5 $2 against GPT-4o $2.50 per million input tokens — a modest gap of 1.3x.
Whole documents in one request: GPT-4o stops at 128,000; Claude Sonnet 5 takes 1,000,000. Past the lower ceiling only Claude Sonnet 5 accepts the request, at any price.
10,000 in, 1,000 out: $0.0350 on GPT-4o, $0.0300 on Claude Sonnet 5. Over 10,000 such calls that is $50.00 in Claude Sonnet 5's favour.
How these counts were arrived at: GPT-4o is counted exactly in your browser; Claude Sonnet 5 is estimated from a characters-per-token ratio. Close, not exact.

GPT-4o details Claude Sonnet 5 details

GPT-4o vs Claude Opus 5

A familiar workhorse against the most capable model on the list.$2.50 vs $5.00 per 1M input

Attribute	GPT-4o	Claude Opus 5
Input, per 1M tokens	$2.50	$5.00
Output, per 1M tokens	$10.00	$25.00
Context window	128,000	1,000,000
Provider	OpenAI	Anthropic
Tokenizer	o200k_base	Anthropic tokenizer
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: GPT-4o $2.50 against Claude Opus 5 $5 per million input tokens — a material gap of twice.
Writing more than you read: GPT-4o prices output at four times its input rate, Claude Opus 5 at 5x — so GPT-4o's lead widens as answers lengthen.
Whole documents in one request: GPT-4o stops at 128,000; Claude Opus 5 takes 1,000,000. Past the lower ceiling only Claude Opus 5 accepts the request, at any price.
10,000 in, 1,000 out: $0.0350 on GPT-4o, $0.0750 on Claude Opus 5. Over 10,000 such calls that is $400.00 in GPT-4o's favour.
How these counts were arrived at: GPT-4o is counted exactly in your browser; Claude Opus 5 is estimated from a characters-per-token ratio. Close, not exact.

GPT-4o details Claude Opus 5 details

Claude Sonnet 5 vs Claude Haiku 4.5

Where the drop from Sonnet to Haiku stops being worth the saving.$2.00 vs $1.00 per 1M input

Attribute	Claude Sonnet 5	Claude Haiku 4.5
Input, per 1M tokens	$2.00	$1.00
Output, per 1M tokens	$10.00	$5.00
Context window	1,000,000	200,000
Provider	Anthropic	Anthropic
Tokenizer	Anthropic tokenizer	Anthropic tokenizer
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: Claude Haiku 4.5 $1 against Claude Sonnet 5 $2 per million input tokens — a material gap of twice.
Whole documents in one request: Claude Haiku 4.5 stops at 200,000; Claude Sonnet 5 takes 1,000,000. Past the lower ceiling only Claude Sonnet 5 accepts the request, at any price.
10,000 in, 1,000 out: $0.0300 on Claude Sonnet 5, $0.0150 on Claude Haiku 4.5. Over 10,000 such calls that is $150.00 in Claude Haiku 4.5's favour.
How these counts were arrived at: Both estimated, and on different ratios (claude-4-7, claude-legacy). Small differences between them are noise.

Claude Sonnet 5 details Claude Haiku 4.5 details

Claude Opus 5 vs Claude Haiku 4.5

The widest price gap between two models from one provider.$5.00 vs $1.00 per 1M input

Attribute	Claude Opus 5	Claude Haiku 4.5
Input, per 1M tokens	$5.00	$1.00
Output, per 1M tokens	$25.00	$5.00
Context window	1,000,000	200,000
Provider	Anthropic	Anthropic
Tokenizer	Anthropic tokenizer	Anthropic tokenizer
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: Claude Haiku 4.5 $1 against Claude Opus 5 $5 per million input tokens — a wide gap of 5x.
Whole documents in one request: Claude Haiku 4.5 stops at 200,000; Claude Opus 5 takes 1,000,000. Past the lower ceiling only Claude Opus 5 accepts the request, at any price.
10,000 in, 1,000 out: $0.0750 on Claude Opus 5, $0.0150 on Claude Haiku 4.5. Over 10,000 such calls that is $600.00 in Claude Haiku 4.5's favour.
How these counts were arrived at: Both estimated, and on different ratios (claude-4-7, claude-legacy). Small differences between them are noise.

Claude Opus 5 details Claude Haiku 4.5 details

GPT-4.1 vs GPT-4o

Two OpenAI models a generation apart, and which one a long prompt favours.$2.00 vs $2.50 per 1M input

Attribute	GPT-4.1	GPT-4o
Input, per 1M tokens	$2.00	$2.50
Output, per 1M tokens	$8.00	$10.00
Context window	1,047,576	128,000
Provider	OpenAI	OpenAI
Tokenizer	o200k_base	o200k_base
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: GPT-4.1 $2 against GPT-4o $2.50 per million input tokens — a modest gap of 1.3x.
Whole documents in one request: GPT-4o stops at 128,000; GPT-4.1 takes 1,047,576. Past the lower ceiling only GPT-4.1 accepts the request, at any price.
10,000 in, 1,000 out: $0.0280 on GPT-4.1, $0.0350 on GPT-4o. Over 10,000 such calls that is $70.00 in GPT-4.1's favour.

GPT-4.1 details GPT-4o details

GPT-4.1 vs GPT-5.6 Terra

Whether the current mid-tier is worth moving to from GPT-4.1.$2.00 vs $2.00 per 1M input

Attribute	GPT-4.1	GPT-5.6 Terra
Input, per 1M tokens	$2.00	$2.00
Output, per 1M tokens	$8.00	$12.00
Context window	1,047,576	1,050,000
Provider	OpenAI	OpenAI
Tokenizer	o200k_base	o200k_base
Long-context tier	none	above 272,000
Retires	no date announced	no date announced

Reading more than you write: Both $2 per million input. Input rate cannot separate this pair.
Writing more than you read: GPT-4.1 prices output at four times its input rate, GPT-5.6 Terra at 6x — so GPT-4.1's lead narrows as answers lengthen.
Whole documents in one request: GPT-4.1 stops at 1,047,576; GPT-5.6 Terra takes 1,050,000. Past the lower ceiling only GPT-5.6 Terra accepts the request, at any price.
Past the long-context threshold: GPT-5.6 Terra above 272,000 bills the whole request higher — $2 and $4. GPT-4.1 stays ahead either side of that line.
10,000 in, 1,000 out: $0.0280 on GPT-4.1, $0.0320 on GPT-5.6 Terra. Over 10,000 such calls that is $40.00 in GPT-4.1's favour.

GPT-4.1 details GPT-5.6 Terra details

GPT-3.5 Turbo vs GPT-5.6 Luna

Where to go when the cheapest old model is switched off.$0.50 vs $0.20 per 1M input

Attribute	GPT-3.5 Turbo	GPT-5.6 Luna
Input, per 1M tokens	$0.50	$0.20
Output, per 1M tokens	$1.50	$1.20
Context window	16,385	1,050,000
Provider	OpenAI	OpenAI
Tokenizer	cl100k_base	o200k_base
Long-context tier	none	above 272,000
Retires	2026-10-23	no date announced

Reading more than you write: GPT-5.6 Luna $0.20 against GPT-3.5 Turbo $0.50 per million input tokens — a material gap of 2.5x.
Writing more than you read: GPT-3.5 Turbo prices output at three times its input rate, GPT-5.6 Luna at 6x — so GPT-5.6 Luna's lead widens as answers lengthen.
Whole documents in one request: GPT-3.5 Turbo stops at 16,385; GPT-5.6 Luna takes 1,050,000. Past the lower ceiling only GPT-5.6 Luna accepts the request, at any price.
Past the long-context threshold: GPT-5.6 Luna above 272,000 bills the whole request higher — $0.40 and $0.50. GPT-5.6 Luna stays ahead either side of that line.
10,000 in, 1,000 out: $0.0065 on GPT-3.5 Turbo, $0.0032 on GPT-5.6 Luna. Over 10,000 such calls that is $33.00 in GPT-5.6 Luna's favour.
One of these is going away: GPT-3.5 Turbo retires 23 October 2026; GPT-5.6 Luna has no date. Removed models fail outright rather than degrade, so past that date price is irrelevant.

GPT-3.5 Turbo details GPT-5.6 Luna details

GPT-4 Turbo vs GPT-4o

The other migration off GPT-4 Turbo, without changing generation.$10.00 vs $2.50 per 1M input

Attribute	GPT-4 Turbo	GPT-4o
Input, per 1M tokens	$10.00	$2.50
Output, per 1M tokens	$30.00	$10.00
Context window	128,000	128,000
Provider	OpenAI	OpenAI
Tokenizer	cl100k_base	o200k_base
Long-context tier	none	none
Retires	2026-10-23	no date announced

Reading more than you write: GPT-4o $2.50 against GPT-4 Turbo $10 per million input tokens — a wide gap of four times.
Writing more than you read: GPT-4 Turbo prices output at three times its input rate, GPT-4o at four times — so GPT-4o's lead widens as answers lengthen.
10,000 in, 1,000 out: $0.1300 on GPT-4 Turbo, $0.0350 on GPT-4o. Over 10,000 such calls that is $950.00 in GPT-4o's favour.
One of these is going away: GPT-4 Turbo retires 23 October 2026; GPT-4o has no date. Removed models fail outright rather than degrade, so past that date price is irrelevant.

GPT-4 Turbo details GPT-4o details

GPT-5.6 Luna vs GPT-4o mini

The two cheapest OpenAI options, where volume makes the gap matter.$0.20 vs $0.15 per 1M input

Attribute	GPT-5.6 Luna	GPT-4o mini
Input, per 1M tokens	$0.20	$0.15
Output, per 1M tokens	$1.20	$0.60
Context window	1,050,000	128,000
Provider	OpenAI	OpenAI
Tokenizer	o200k_base	o200k_base
Long-context tier	above 272,000	none
Retires	no date announced	no date announced

Reading more than you write: GPT-4o mini $0.15 against GPT-5.6 Luna $0.20 per million input tokens — a modest gap of 1.3x.
Writing more than you read: GPT-5.6 Luna prices output at 6x its input rate, GPT-4o mini at four times — so GPT-4o mini's lead widens as answers lengthen.
Whole documents in one request: GPT-4o mini stops at 128,000; GPT-5.6 Luna takes 1,050,000. Past the lower ceiling only GPT-5.6 Luna accepts the request, at any price.
Past the long-context threshold: GPT-5.6 Luna above 272,000 bills the whole request higher — $0.15 and $0.40. GPT-4o mini stays ahead either side of that line.
10,000 in, 1,000 out: $0.0032 on GPT-5.6 Luna, $0.0021 on GPT-4o mini. Over 10,000 such calls that is $11.00 in GPT-4o mini's favour.

GPT-5.6 Luna details GPT-4o mini details

Gemini 3.1 Pro vs GPT-5.6 Terra

Two flagship-adjacent models priced identically on input, and where the bills diverge.$2.00 vs $2.00 per 1M input

Attribute	Gemini 3.1 Pro	GPT-5.6 Terra
Input, per 1M tokens	$2.00	$2.00
Output, per 1M tokens	$12.00	$12.00
Context window	1,000,000	1,050,000
Provider	Google	OpenAI
Tokenizer	Google tokenizer	o200k_base
Long-context tier	above 200,000	above 272,000
Retires	no date announced	no date announced

Reading more than you write: Both $2 per million input. Input rate cannot separate this pair.
Whole documents in one request: Gemini 3.1 Pro stops at 1,000,000; GPT-5.6 Terra takes 1,050,000. Past the lower ceiling only GPT-5.6 Terra accepts the request, at any price.
Past the long-context threshold: Gemini 3.1 Pro above 200,000, GPT-5.6 Terra above 272,000 bills the whole request higher — $4 and $4. Gemini 3.1 Pro stays ahead either side of that line.
One rate here is provisional: Gemini 3.1 Pro ships as a preview, so $2 is not a committed rate.
How these counts were arrived at: GPT-5.6 Terra is counted exactly in your browser; Gemini 3.1 Pro is estimated from a characters-per-token ratio. Close, not exact.

Gemini 3.1 Pro details GPT-5.6 Terra details

Gemini 3.1 Pro vs Claude Opus 5

Google's flagship against Anthropic's, including the long-prompt surcharge only one of them charges.$2.00 vs $5.00 per 1M input

Attribute	Gemini 3.1 Pro	Claude Opus 5
Input, per 1M tokens	$2.00	$5.00
Output, per 1M tokens	$12.00	$25.00
Context window	1,000,000	1,000,000
Provider	Google	Anthropic
Tokenizer	Google tokenizer	Anthropic tokenizer
Long-context tier	above 200,000	none
Retires	no date announced	no date announced

Reading more than you write: Gemini 3.1 Pro $2 against Claude Opus 5 $5 per million input tokens — a material gap of 2.5x.
Writing more than you read: Gemini 3.1 Pro prices output at 6x its input rate, Claude Opus 5 at 5x — so Gemini 3.1 Pro's lead widens as answers lengthen.
Past the long-context threshold: Gemini 3.1 Pro above 200,000 bills the whole request higher — $4 and $5. Gemini 3.1 Pro stays ahead either side of that line.
10,000 in, 1,000 out: $0.0320 on Gemini 3.1 Pro, $0.0750 on Claude Opus 5. Over 10,000 such calls that is $430.00 in Gemini 3.1 Pro's favour.
One rate here is provisional: Gemini 3.1 Pro ships as a preview, so $2 is not a committed rate.
How these counts were arrived at: Both estimated, and on different ratios (gemini, claude-4-7). Small differences between them are noise.

Gemini 3.1 Pro details Claude Opus 5 details

Gemini 3.1 Pro vs GPT-5.6 Sol

The two most capable models here, and what the flagship tier costs at each vendor.$2.00 vs $5.00 per 1M input

Attribute	Gemini 3.1 Pro	GPT-5.6 Sol
Input, per 1M tokens	$2.00	$5.00
Output, per 1M tokens	$12.00	$30.00
Context window	1,000,000	1,050,000
Provider	Google	OpenAI
Tokenizer	Google tokenizer	o200k_base
Long-context tier	above 200,000	above 272,000
Retires	no date announced	no date announced

Reading more than you write: Gemini 3.1 Pro $2 against GPT-5.6 Sol $5 per million input tokens — a material gap of 2.5x.
Whole documents in one request: Gemini 3.1 Pro stops at 1,000,000; GPT-5.6 Sol takes 1,050,000. Past the lower ceiling only GPT-5.6 Sol accepts the request, at any price.
Past the long-context threshold: Gemini 3.1 Pro above 200,000, GPT-5.6 Sol above 272,000 bills the whole request higher — $4 and $10. Gemini 3.1 Pro stays ahead either side of that line.
10,000 in, 1,000 out: $0.0320 on Gemini 3.1 Pro, $0.0800 on GPT-5.6 Sol. Over 10,000 such calls that is $480.00 in Gemini 3.1 Pro's favour.
One rate here is provisional: Gemini 3.1 Pro ships as a preview, so $2 is not a committed rate.
How these counts were arrived at: GPT-5.6 Sol is counted exactly in your browser; Gemini 3.1 Pro is estimated from a characters-per-token ratio. Close, not exact.

Gemini 3.1 Pro details GPT-5.6 Sol details

Gemini 3.6 Flash vs GPT-4o

The fast tier from each provider, which is where most everyday API traffic actually goes.$1.50 vs $2.50 per 1M input

Attribute	Gemini 3.6 Flash	GPT-4o
Input, per 1M tokens	$1.50	$2.50
Output, per 1M tokens	$7.50	$10.00
Context window	1,000,000	128,000
Provider	Google	OpenAI
Tokenizer	Google tokenizer	o200k_base
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: Gemini 3.6 Flash $1.50 against GPT-4o $2.50 per million input tokens — a modest gap of 1.7x.
Writing more than you read: Gemini 3.6 Flash prices output at 5x its input rate, GPT-4o at four times — so Gemini 3.6 Flash's lead widens as answers lengthen.
Whole documents in one request: GPT-4o stops at 128,000; Gemini 3.6 Flash takes 1,000,000. Past the lower ceiling only Gemini 3.6 Flash accepts the request, at any price.
10,000 in, 1,000 out: $0.0225 on Gemini 3.6 Flash, $0.0350 on GPT-4o. Over 10,000 such calls that is $125.00 in Gemini 3.6 Flash's favour.
How these counts were arrived at: GPT-4o is counted exactly in your browser; Gemini 3.6 Flash is estimated from a characters-per-token ratio. Close, not exact.

Gemini 3.6 Flash details GPT-4o details

Gemini 3.6 Flash vs Claude Sonnet 5

The mid-tier choice between Google and Anthropic.$1.50 vs $2.00 per 1M input

Attribute	Gemini 3.6 Flash	Claude Sonnet 5
Input, per 1M tokens	$1.50	$2.00
Output, per 1M tokens	$7.50	$10.00
Context window	1,000,000	1,000,000
Provider	Google	Anthropic
Tokenizer	Google tokenizer	Anthropic tokenizer
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: Gemini 3.6 Flash $1.50 against Claude Sonnet 5 $2 per million input tokens — a modest gap of 1.3x.
10,000 in, 1,000 out: $0.0225 on Gemini 3.6 Flash, $0.0300 on Claude Sonnet 5. Over 10,000 such calls that is $75.00 in Gemini 3.6 Flash's favour.
How these counts were arrived at: Both estimated, and on different ratios (gemini, claude-4-7). Small differences between them are noise.

Gemini 3.6 Flash details Claude Sonnet 5 details

Gemini 3.1 Flash-Lite vs GPT-5.6 Luna

The cheapest current model from each of Google and OpenAI, for high-volume work.$0.25 vs $0.20 per 1M input

Attribute	Gemini 3.1 Flash-Lite	GPT-5.6 Luna
Input, per 1M tokens	$0.25	$0.20
Output, per 1M tokens	$1.50	$1.20
Context window	1,000,000	1,050,000
Provider	Google	OpenAI
Tokenizer	Google tokenizer	o200k_base
Long-context tier	none	above 272,000
Retires	no date announced	no date announced

Reading more than you write: GPT-5.6 Luna $0.20 against Gemini 3.1 Flash-Lite $0.25 per million input tokens — a modest gap of 1.3x.
Whole documents in one request: Gemini 3.1 Flash-Lite stops at 1,000,000; GPT-5.6 Luna takes 1,050,000. Past the lower ceiling only GPT-5.6 Luna accepts the request, at any price.
Past the long-context threshold: The ranking inverts (GPT-5.6 Luna above 272,000): GPT-5.6 Luna wins below the line, Gemini 3.1 Flash-Lite above it at $0.25 against $0.40. Judging this pair on headline rates alone gives the wrong answer for large prompts.
10,000 in, 1,000 out: $0.004 on Gemini 3.1 Flash-Lite, $0.0032 on GPT-5.6 Luna. Over 10,000 such calls that is $8.00 in GPT-5.6 Luna's favour.
How these counts were arrived at: GPT-5.6 Luna is counted exactly in your browser; Gemini 3.1 Flash-Lite is estimated from a characters-per-token ratio. Close, not exact.

Gemini 3.1 Flash-Lite details GPT-5.6 Luna details

Gemini 3.1 Flash-Lite vs Claude Haiku 4.5

Budget against budget, where the per-token gap decides the whole bill.$0.25 vs $1.00 per 1M input

Attribute	Gemini 3.1 Flash-Lite	Claude Haiku 4.5
Input, per 1M tokens	$0.25	$1.00
Output, per 1M tokens	$1.50	$5.00
Context window	1,000,000	200,000
Provider	Google	Anthropic
Tokenizer	Google tokenizer	Anthropic tokenizer
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: Gemini 3.1 Flash-Lite $0.25 against Claude Haiku 4.5 $1 per million input tokens — a wide gap of four times.
Writing more than you read: Gemini 3.1 Flash-Lite prices output at 6x its input rate, Claude Haiku 4.5 at 5x — so Gemini 3.1 Flash-Lite's lead widens as answers lengthen.
Whole documents in one request: Claude Haiku 4.5 stops at 200,000; Gemini 3.1 Flash-Lite takes 1,000,000. Past the lower ceiling only Gemini 3.1 Flash-Lite accepts the request, at any price.
10,000 in, 1,000 out: $0.004 on Gemini 3.1 Flash-Lite, $0.0150 on Claude Haiku 4.5. Over 10,000 such calls that is $110.00 in Gemini 3.1 Flash-Lite's favour.
How these counts were arrived at: Both estimated, and on different ratios (gemini, claude-legacy). Small differences between them are noise.

Gemini 3.1 Flash-Lite details Claude Haiku 4.5 details

Gemini 3.1 Pro vs Gemini 3.6 Flash

The two Gemini models people actually search for, and how far Flash gets you.$2.00 vs $1.50 per 1M input

Attribute	Gemini 3.1 Pro	Gemini 3.6 Flash
Input, per 1M tokens	$2.00	$1.50
Output, per 1M tokens	$12.00	$7.50
Context window	1,000,000	1,000,000
Provider	Google	Google
Tokenizer	Google tokenizer	Google tokenizer
Long-context tier	above 200,000	none
Retires	no date announced	no date announced

Reading more than you write: Gemini 3.6 Flash $1.50 against Gemini 3.1 Pro $2 per million input tokens — a modest gap of 1.3x.
Writing more than you read: Gemini 3.1 Pro prices output at 6x its input rate, Gemini 3.6 Flash at 5x — so Gemini 3.6 Flash's lead widens as answers lengthen.
Past the long-context threshold: Gemini 3.1 Pro above 200,000 bills the whole request higher — $1.50 and $4. Gemini 3.6 Flash stays ahead either side of that line.
10,000 in, 1,000 out: $0.0320 on Gemini 3.1 Pro, $0.0225 on Gemini 3.6 Flash. Over 10,000 such calls that is $95.00 in Gemini 3.6 Flash's favour.
One rate here is provisional: Gemini 3.1 Pro ships as a preview, so $2 is not a committed rate.

Gemini 3.1 Pro details Gemini 3.6 Flash details

Gemini 3.6 Flash vs Gemini 2.5 Flash

Whether the newer Flash earns five times the price of the one it replaced.$1.50 vs $0.30 per 1M input

Attribute	Gemini 3.6 Flash	Gemini 2.5 Flash
Input, per 1M tokens	$1.50	$0.30
Output, per 1M tokens	$7.50	$2.50
Context window	1,000,000	1,000,000
Provider	Google	Google
Tokenizer	Google tokenizer	Google tokenizer
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: Gemini 2.5 Flash $0.30 against Gemini 3.6 Flash $1.50 per million input tokens — a wide gap of 5x.
Writing more than you read: Gemini 3.6 Flash prices output at 5x its input rate, Gemini 2.5 Flash at 8.3x — so Gemini 2.5 Flash's lead widens as answers lengthen.
10,000 in, 1,000 out: $0.0225 on Gemini 3.6 Flash, $0.0055 on Gemini 2.5 Flash. Over 10,000 such calls that is $170.00 in Gemini 2.5 Flash's favour.

Gemini 3.6 Flash details Gemini 2.5 Flash details

Gemini 3.1 Flash-Lite vs Gemini 2.5 Flash

Two Gemini models priced within a nickel of each other, a generation apart.$0.25 vs $0.30 per 1M input

Attribute	Gemini 3.1 Flash-Lite	Gemini 2.5 Flash
Input, per 1M tokens	$0.25	$0.30
Output, per 1M tokens	$1.50	$2.50
Context window	1,000,000	1,000,000
Provider	Google	Google
Tokenizer	Google tokenizer	Google tokenizer
Long-context tier	none	none
Retires	no date announced	no date announced

Reading more than you write: Gemini 3.1 Flash-Lite $0.25 against Gemini 2.5 Flash $0.30 per million input tokens — near-parity of 1.2x.
Writing more than you read: Gemini 3.1 Flash-Lite prices output at 6x its input rate, Gemini 2.5 Flash at 8.3x — so Gemini 3.1 Flash-Lite's lead widens as answers lengthen.
10,000 in, 1,000 out: $0.004 on Gemini 3.1 Flash-Lite, $0.0055 on Gemini 2.5 Flash. Over 10,000 such calls that is $15.00 in Gemini 3.1 Flash-Lite's favour.

Gemini 3.1 Flash-Lite details Gemini 2.5 Flash details

What this page does not tell you. Which model produces better answers. That depends on your task, and this site has no benchmark data — everything above is drawn from published pricing and specifications. Compare quality by running your own prompts through both.

These pairs are chosen rather than generated. Every combination of the models here would be more than fifty pairs, most of them answering a question nobody asks — so any pair not listed can be compared directly in the full table, which prices all of them against the same prompt at once.