Price history

LLM API Price Changes, 2024–2026

47 verified price changes from 6 providers since 2024. Older changes link archived copies of the official pricing page from before and after; newer ones were caught by our price monitor and link the official page.

Changes by provider

Provider	Changes	Cuts	Increases
Google Gemini	15	9	5
OpenAI	10	5	0
DeepSeek	9	4	2
Mistral AI	6	5	1
Anthropic	4	4	0
xAI	3	2	0

Increases include ended discounts. Other changes are new cached-input prices, aliases moved to a new snapshot and preview models repriced at launch.

2026

Date	Model	Change	Price per 1M tokens	Archived pages
Oct 31, 2026first seen by our monitor	Mistral Large 4Mistral AI	Price cut	Input: $1.36 → $0.70 Output: $4.18 → $2.10 Cached input: $0.14 → $0.07	Official page Monitor entry
Oct 21, 2026first seen by our monitor	Mistral Large 4Mistral AI	Price increase	Input: $0.68 → $1.36 Output: $2.09 → $4.18 Cached input: $0.07 → $0.14	Official page Monitor entry
Oct 7, 2026	Claude Sonnet 5.5Anthropic	Price cut	Cache read: $0.20 → $0.10	Official page Announcement Model page
Aug 29, 2026first seen by our monitor	Gemini 3.6 FlashGoogle Gemini	Price cut	Input: $1.50 → $0.75 Output: $7.50 → $3.75 Cached input: $0.15 → $0.075	Official page Monitor entry
Aug 22, 2026first seen by our monitor	GPT-5.6 SolOpenAI	Price cut	Input: $5.00 → $4.00 Output: $30.00 → $20.00 Cached input: $0.50 → $0.40	Official page Monitor entry
Jul 30, 2026	GPT-5.6 LunaOpenAI	Price cut	Input: $1.00 → $0.20 Cached input: $0.10 → $0.02 Cache write: $1.25 → $0.25 Output: $6.00 → $1.20 Long-context input: $2.00 → $0.40 Long-context cached input: $0.20 → $0.04 Long-context cache write: $2.50 → $0.50 Long-context output: $9.00 → $1.80	Before After Announcement
Jul 30, 2026	GPT-5.6 TerraOpenAI	Price cut	Input: $2.50 → $2.00 Cached input: $0.25 → $0.20 Cache write: $3.125 → $2.50 Output: $15.00 → $12.00 Long-context input: $5.00 → $4.00 Long-context cached input: $0.50 → $0.40 Long-context cache write: $6.25 → $5.00 Long-context output: $22.50 → $18.00	Before After Announcement
Jul 27, 2026first seen by our monitor	Mistral Large 3Mistral AI	Price cut	Input: $2.00 → $0.50 Output: $6.00 → $1.50 Cached input: new → $0.05	Official page Monitor entry
Jul 27, 2026first seen by our monitor	Grok 4.3xAI	Cached price added	Cached input: new → $0.20	Official page Model page
May 31, 2026	deepseek-v4-proDeepSeek	Price cut	Cache-hit input: $0.0145 → $0.003625 Input: $1.74 → $0.435 Output: $3.48 → $0.87	Before After
May 1, 2026 – May 6, 2026between archived captures	grok-4.20-non-reasoningxAI	Price cut	Input: $2.00 → $1.25 Output: $6.00 → $2.50	Before After
May 1, 2026 – May 6, 2026between archived captures	grok-4.20-reasoningxAI	Price cut	Input: $2.00 → $1.25 Output: $6.00 → $2.50	Before After
Apr 26, 2026	deepseek-v4-flashDeepSeek	Price cut	Cache-hit input: $0.028 → $0.0028	Before After
Apr 26, 2026	deepseek-v4-proDeepSeek	Price cut	Cache-hit input: $0.145 → $0.0145 Cache-hit input (discounted): $0.0363 → $0.003625	Before After
Apr 24, 2026 – Apr 25, 2026between archived captures	deepseek-v4-proDeepSeek	Discount started	Cache-hit input: $0.145 → $0.0363 Input: $1.74 → $0.435 Output: $3.48 → $0.87	Before After
Apr 23, 2026 – Apr 24, 2026between archived captures	deepseek-chat / deepseek-reasonerDeepSeek	Alias moved to a new snapshot	Input: $0.28 → $0.14 Output: $0.42 → $0.28	Before After Announcement
Mar 13, 2026	Claude Opus 4.6Anthropic	Price cut	Input over 200K: $10.00 → $5.00 Output over 200K: $37.50 → $25.00	Before After Announcement
Mar 13, 2026	Claude Sonnet 4.6Anthropic	Price cut	Input over 200K: $6.00 → $3.00 Output over 200K: $22.50 → $15.00	Before After Announcement

2025

Date	Model	Change	Price per 1M tokens	Archived pages
Oct 24, 2025between archived captures	Gemini 2.5 Flash-LiteGoogle Gemini	Price cut	Cached input: $0.025 → $0.01 Cached audio input: $0.125 → $0.03	Before After
Oct 24, 2025between archived captures	Gemini 2.5 Flash-Lite Preview (09-2025)Google Gemini	Price cut	Cached input: $0.025 → $0.01 Cached audio input: $0.125 → $0.03	Before After
Oct 16, 2025between archived captures	Gemini 2.5 Flash Preview (09-2025)Google Gemini	Price cut	Cached input: $0.0375 → $0.03 Cached audio input: $0.25 → $0.10	Before After
Oct 5, 2025 – Oct 9, 2025between archived captures	Gemini 2.5 FlashGoogle Gemini	Price cut	Cached input: $0.075 → $0.03 Cached audio input: $0.25 → $0.10	Before After
Oct 5, 2025 – Oct 9, 2025between archived captures	Gemini 2.5 ProGoogle Gemini	Price cut	Cached input: $0.31 → $0.125 Cached input over 200K: $0.625 → $0.25	Before After
Sep 29, 2025	deepseek-chat / deepseek-reasoner (DeepSeek-V3.2-Exp)DeepSeek	Price cut	Cache-hit input: $0.07 → $0.028 Input: $0.56 → $0.28 Output: $1.68 → $0.42	Before After Announcement
Sep 3, 2025 – Sep 7, 2025between archived captures	Gemini 2.0 FlashGoogle Gemini	Price increase	Cached input: $0.0125 → $0.025 Cached audio input: $0.0875 → $0.175	Before After
Sep 3, 2025 – Sep 7, 2025between archived captures	Gemini 2.5 FlashGoogle Gemini	Price increase	Cached input: $0.0375 → $0.075 Cached audio input: $0.125 → $0.25	Before After
Sep 3, 2025 – Sep 7, 2025between archived captures	Gemini 2.5 Flash-LiteGoogle Gemini	Price increase	Cached input: $0.0125 → $0.025 Cached audio input: $0.0625 → $0.125	Before After
Sep 3, 2025 – Sep 7, 2025between archived captures	Gemini 2.5 ProGoogle Gemini	Price increase	Cached input: $0.155 → $0.31 Cached input over 200K: $0.3125 → $0.625	Before After
Sep 5, 2025	deepseek-chat (DeepSeek-V3.1 non-thinking)DeepSeek	Price increase	Input: $0.27 → $0.56 Output: $1.10 → $1.68	Before After Announcement
Sep 5, 2025	deepseek-reasoner (DeepSeek-V3.1 thinking)DeepSeek	Repriced	Cache-hit input: $0.14 → $0.07 Input: $0.55 → $0.56 Output: $2.19 → $1.68	Before After Announcement
Jun 17, 2025	Gemini 2.5 FlashGoogle Gemini	Preview to stable reprice	Input: $0.15 → $0.30 Output (non-thinking): $0.60 → $2.50 Output (thinking): $3.50 → $2.50	Before After Announcement
Jun 10, 2025	o3OpenAI	Price cut	Input: $10.00 → $2.00 Cached input: $2.50 → $0.50 Output: $40.00 → $8.00	Before After Announcement
Feb 8, 2025	deepseek-chat (DeepSeek-V3)DeepSeek	Discount ended	Cache-hit input: $0.014 → $0.07 Input: $0.14 → $0.27 Output: $0.28 → $1.10	Before After Announcement
Jan 27, 2025 – Feb 1, 2025between archived captures	o1-miniOpenAI	Price cut	Input: $3.00 → $1.10 Cached input: $1.50 → $0.55 Output: $12.00 → $4.40	Before After

2024

Date	Model	Change	Price per 1M tokens	Archived pages
Dec 3, 2024between archived captures	Claude 3.5 HaikuAnthropic	Price cut	Input: $1.00 → $0.80 Output: $5.00 → $4.00 Cache write: $1.25 → $1.00 Cache read: $0.10 → $0.08	Before After
Oct 2, 2024 – Oct 3, 2024between archived captures	gpt-4oOpenAI	Alias moved to a new snapshot	Input: $5.00 → $2.50 Output: $15.00 → $10.00 Batch input: $2.50 → $1.25	Before After
Oct 1, 2024	Gemini 1.5 ProGoogle Gemini	Price cut	Input: $3.50 → $1.25 Output: $10.50 → $5.00 Long-context input: $7.00 → $2.50 Long-context output: $21.00 → $10.00 Cached input: $0.875 → $0.3125	Before After Announcement
Oct 1, 2024	gpt-4o-2024-08-06OpenAI	Cached price added	Cached input: new → $1.25	Before After Announcement
Oct 1, 2024	gpt-4o-miniOpenAI	Cached price added	Cached input: new → $0.075	Before After Announcement
Oct 1, 2024	o1-miniOpenAI	Cached price added	Cached input: new → $1.50	Before After Announcement
Oct 1, 2024	o1-previewOpenAI	Cached price added	Cached input: new → $7.50	Before After Announcement
Sep 17, 2024	CodestralMistral AI	Price cut	Input: $1.00 → $0.20 Output: $3.00 → $0.60	Before After Announcement
Sep 17, 2024	Mistral Large 2Mistral AI	Price cut	Input: $3.00 → $2.00 Output: $9.00 → $6.00	Before After Announcement
Sep 17, 2024	Mistral NeMoMistral AI	Price cut	Input: $0.30 → $0.15 Output: $0.30 → $0.15	Before After Announcement
Aug 12, 2024	Gemini 1.5 FlashGoogle Gemini	Price cut	Input: $0.35 → $0.075 Long-context input: $0.70 → $0.15 Output: $1.05 → $0.30 Long-context output: $2.10 → $0.60 Cached input: $0.0875 → $0.0188	Before After Announcement
Jun 15, 2024 – Jun 20, 2024between archived captures	Gemini 1.5 ProGoogle Gemini	Price cut	Cached input: $1.75 → $0.875 Long-context cached input: $3.50 → $1.75	Before After
May 16, 2024 – May 18, 2024between archived captures	Gemini 1.5 FlashGoogle Gemini	Price increase	Output: $0.53 → $1.05 Long-context output: $1.05 → $2.10	Before After

Announced, not yet in effect

Future prices a provider has already published. They join the tables above on the day they take effect.

Dec 1, 2026: From Dec 1, 2026, Gemini 4 Argon is scheduled to change input $2.00 → $4.00 (+100%) and output $10.00 → $20.00 (+100%) per 1M tokens, as announced by Google Gemini. Monitor entry · Model page
Jan 1, 2027: From Jan 1, 2027, Gemini 3.6 Flash is scheduled to change input $0.75 → $1.50 (+100%) and output $3.75 → $7.50 (+100%) per 1M tokens, as announced by Google Gemini. Monitor entry · Model page
Jan 1, 2027: From Jan 1, 2027, Gemini 3.8 Flash is scheduled to change input $0.75 → $1.50 (+100%) and output $3.75 → $7.50 (+100%) per 1M tokens, as announced by Google Gemini. Monitor entry · Model page

How each change was checked

The list has two sources. 40 older changes were rebuilt from archived pricing pages, and 7 newer ones were caught by our own model price monitor, which checks the official pricing pages and records a change once it is confirmed there. Monitor rows link the official page, the announcement when there is one, and the monitor entry.

Each archived row was rebuilt from Internet Archive captures of the provider's own pricing page. “Before” opens the last capture that still shows the old price and “After” the first capture that shows the new one. When the provider published an announcement or changelog entry, it is linked too, and its date is the date shown. Without one, the change happened somewhere between the two captures.

Third-party price lists were only used to find candidates. A candidate was added only after two official captures showed the change. New models launched at a lower price than an older one are not counted as price changes.

What was checked for each provider, and the gaps that remain, are in the dataset documentation. The JSON file has the quoted page text for every capture.

Current prices and new models

New model launches, retirements and successor models are not price changes, so they are not listed here. They are in the model price monitor, with a feed you can subscribe to. For today's prices, use the model pricing comparison or the model pricing pages.

What the list shows about timing, such as how soon after launch a cut tends to arrive and where the increases land, is in do LLM API prices go down?

Frequently asked questions

Do LLM API prices usually go down?

Mostly. Of the 47 changes listed here, 29 are price cuts and 8 are increases or ended discounts. The largest input price cut was Mistral AI Codestral, down 80% (Sep 17, 2024). Most cuts arrive together with, or soon after, a newer model.

Which API prices went up?

Mistral AI Mistral Large 4 (Oct 21, 2026); Google Gemini Gemini 2.0 Flash (Sep 3, 2025 – Sep 7, 2025); Google Gemini Gemini 2.5 Flash (Sep 3, 2025 – Sep 7, 2025); Google Gemini Gemini 2.5 Flash-Lite (Sep 3, 2025 – Sep 7, 2025); Google Gemini Gemini 2.5 Pro (Sep 3, 2025 – Sep 7, 2025); DeepSeek deepseek-chat (DeepSeek-V3.1 non-thinking) (Sep 5, 2025); DeepSeek deepseek-chat (DeepSeek-V3) (Feb 8, 2025); Google Gemini Gemini 1.5 Flash (May 16, 2024 – May 18, 2024). Several of these are cached-input or Batch rates, not the headline input and output prices.

How do you know when each change happened?

Older changes link two captures of the provider's official pricing page in the Internet Archive: the last one that still shows the old price and the first one that shows the new price. Newer changes were caught by our own price monitor, which checks the official pricing pages, and link the page itself. When the provider announced a date, that date is shown; otherwise it is the capture window or the day the monitor first saw the new price.

Is this every price change since 2024?

No. It covers Google Gemini 15, OpenAI 10, DeepSeek 9, Mistral AI 6, Anthropic 4, xAI 3 changes. Older ones are included only when two readable archived pages confirm them. Some pricing pages are rendered by JavaScript and have no readable archive, and image, audio-only and embedding prices are not included. The dataset documentation lists what was checked for each provider and what is still open.

Can I download this data?

Yes. The archived changes, with every quoted page and link, are a JSON file in the public AI API pricing dataset on GitHub. Newer changes are in the same dataset's price change events and in the monitor's Atom feed.