AI API Pricing November 2026: What Changed This Month
Claude Haiku 5.5 launched at $0.10/$0.50, GPT-6.1 Terra at $1/$5 and Gemini 4 Argon at an introductory $2/$10, while Mistral cut Large 4 to $0.70/$2.10. Every AI API price change since 5 October 2026, and the deadlines to watch.
API price changes in November 2026 so far
Our daily price check adds each API price change, launch and retirement here as it happens, newest first. The full record is on our price changes page.
No API prices have changed yet in November 2026. New changes appear here the day our check logs them.
On this page
Our AI API pricing November 2026 review starts with lower prices almost everywhere. Since 5 October, Claude Haiku 5.5 has launched at $0.10/$0.50 per million input/output tokens, GPT-6.1 Terra at $1/$5 and Gemini 4 Argon at $2/$10.
Mistral cut Large 4 to $0.70/$2.10. The next scheduled rise is Gemini 4 Argon, which doubles to $4/$20 on 1 December.
We checked every price in this post against the vendors' own pricing pages on 3 Nov 2026. Google's pages could not be read that day, so its figures come from our rate card, whose prices were last checked on 30 Oct 2026.
This review covers every change since our October 2026 AI API pricing roundup, model by model, and what each one means for your bill. The list above this review tracks every change this month as it lands.
AI API Pricing November 2026: The Short Version
- Claude Haiku 5.5 launched on 7 Oct at $0.10/$0.50 for prompts up to 100,000 tokens. That is a tenth of Claude Haiku 4.5's $1/$5.
- Anthropic halved Claude Sonnet 5.5's cache reads on 7 Oct, from $0.20 to $0.10 per million tokens.
- Gemini 4 Argon went GA on 13 Oct at an introductory $2/$10. Its $4/$20 list price starts on 1 Dec.
- GPT-6.1 Terra launched on 26 Oct at $1/$5, half the price of GPT-6.1 Sol.
- Mistral Large 4 arrived on 6 Oct in public preview, at a $1.36/$4.18 list price after a two-week launch discount. By 30 Oct, Mistral listed it at $0.70/$2.10.
- Mistral deprecated Large 3 and Medium 3.5 on 31 Oct. Both leave its API on 30 Apr 2027.
- OpenAI shut down o4-mini on 23 Oct, and Google began routing Gemini 3.5 Flash to Gemini 3.6 Flash on 8 Oct.
Every LLM API Price Change Since 5 October
Our October roundup was checked on 5 Oct 2026. These are the LLM API price changes we logged after that, in date order.
| Date | Provider | What Changed | Rate Now (per MTok) |
|---|---|---|---|
| 6 Oct | Mistral | Mistral Large 4 launches in public preview, with a two-week launch discount | $1.36/$4.18 list at launch |
| 7 Oct | Anthropic | Claude Haiku 5.5 launches at a tenth of Haiku 4.5's price | $0.10/$0.50 up to 100K tokens |
| 7 Oct | Anthropic | Claude Sonnet 5.5 cache reads halved | $0.10 cache reads |
| 8 Oct | Gemini 3.5 Flash deprecated, with its requests routed to Gemini 3.6 Flash | $0.75/$3.75 on 3.6 Flash | |
| 13 Oct | Gemini 4 Argon goes GA at an introductory rate | $2/$10 until 30 Nov | |
| 20 Oct | OpenAI | 12-hour prompt caching added | 12-hour writes at 2.5x input |
| 23 Oct | OpenAI | o4-mini API shut down | Removed |
| 26 Oct | OpenAI | GPT-6.1 Terra launches | $1/$5 |
| By 26 Oct | xAI | 20% batch discount listed for Grok 4.7, 4.6 and 4.5 | $1.60/$4.80 batch on Grok 4.7 |
| By 26 Oct | Meta | Muse Spark Contributor mode cap raised from 100 to 150 requests a minute | $0.10/$0.20 in Contributor mode |
| By 30 Oct | Mistral | Mistral Large 4 cut by about half | $0.70/$2.10 |
| 31 Oct | Mistral | Large 3 and Medium 3.5 deprecated, retiring on 30 Apr 2027 | $0.50/$1.50 and $1.50/$7.50 until then |
How We Checked These Prices
Every figure in this AI API pricing November 2026 review comes from the vendor's own pricing page, model page, changelog or deprecation table. None comes from a tracker or a summary.
Mistral, xAI and Meta changed their pages without dating the change. For those, the table says "by" the day we first saw it.
We count list rates, cache and batch rates, long-prompt thresholds and time-of-day billing as price changes. Every change we log, with its date, is on our AI API price changes page.
Claude Haiku 5.5 vs GPT-6 Luna: Same Price, Until 100K Tokens
Anthropic launched Claude Haiku 5.5 on 7 Oct 2026. It calls it "our most capable model tuned for high-volume and latency-sensitive work". It has a 1M-token context window and 128K max output.
For prompts up to 100,000 tokens, Haiku 5.5 costs $0.10/$0.50, with cache reads at $0.01. That is exactly GPT-6 Luna's price. So the cheapest Claude model we track now ties OpenAI's cheapest current-generation model on short prompts.
A Tenth of Haiku 4.5's Price
Claude Haiku 4.5 costs $1/$5, so Haiku 5.5 is a tenth of its price on short prompts. On our standard workload of 300 million input and 50 million output tokens a month, Haiku 4.5 costs $550 and Haiku 5.5 costs $55.
The token count is not quite like for like. Anthropic says the newer tokenizer on Claude 4.7 and later models "produces approximately 30% more tokens for the same text". Even with 30% more tokens, Haiku 5.5 would cost about $71.50 for that job, a saving of roughly 87%.
Haiku 5.5 is cheaper at every prompt length. Above 100,000 tokens it bills $0.50/$2.50, half Haiku 4.5's rate. Anthropic's deprecations table still lists Haiku 4.5 as active, so nothing forces a move.
The work is in the code. Anthropic warns that "Code written for Claude Haiku 4.5 can break on Claude Haiku 5.5". Our migration guides cover the usual changes.
Where Haiku 5.5 Costs Five Times More
The tie ends at 100,000 tokens. A Haiku 5.5 request with a prompt over that size pays $0.50/$2.50 for the whole request, not just the excess.
Anthropic's pricing documentation counts "all of its input tokens, including cache reads and cache writes" toward the threshold. A large cached system prompt can push a short question over the line.
GPT-6 Luna keeps $0.10/$0.50 up to 272K tokens and charges $0.20/$0.75 above that. Between 100K and 272K tokens, Haiku 5.5 costs five times as much as Luna on both input and output.
Above 272K, it still costs 2.5 times Luna's input rate and more than three times its output rate.
So Claude Haiku 5.5 vs GPT-6 Luna comes down to prompt size. For short classification, extraction and routing jobs, the API pricing is identical, so pick on quality.
For retrieval jobs that send long documents, Luna is far cheaper. Both halve to $0.05/$0.25 on the Batch API for short prompts.
Sonnet 5.5's Cache Reads Halved
On the same day, Anthropic cut prompt cache reads on Claude Sonnet 5.5 from $0.20 to $0.10 per million tokens, or 0.05x the input price. Cache writes and the $2/$10 token rates did not change. Sonnet 5.5 now matches GPT-6.1 Sol on input, output and cached input.
With 80% of input served from cache, our standard workload costs $644 a month on either model, before cache writes. In October, at the old cache rate, Sonnet 5.5 cost $668 for the same job.
The bigger gap is still long prompts. Sonnet 5.5 bills its full 1M window at standard rates, while GPT-6.1 Sol rises to $4/$15 above 272K tokens.
See our Claude API pricing guide for the full line-up, and our October 2026 Claude releases roundup for the rest of Anthropic's month.
GPT-6.1 Terra Pricing: OpenAI's New $1/$5 Middle
OpenAI released GPT-6.1 Terra on 26 Oct 2026. It describes it as "a fast, efficient model for visual reasoning, precise computer use, and clear, natural writing". It costs $1/$5, half GPT-6.1 Sol's $2/$10. Cached input is $0.05, or 5% of the input price.
Like the rest of the GPT-6 family, GPT-6.1 Terra pricing has a long-context band. Above 272K input tokens, the whole request bills at $2/$7.50. The context window is 1.05M tokens, with 128K max output, and tool calling requires the Responses API.
GPT-6.1 Terra also undercuts the older GPT-5.6 Terra, which costs $2/$12. On our standard workload, that is $550 a month against $1,200, a 54% saving.
This matters if you left o4-mini in October, because OpenAI's recommended replacement for o4-mini is GPT-5.6 Terra, not the newer model. Our OpenAI API pricing guide lists every GPT tier.
12-Hour Prompt Caching Arrives
On 20 Oct, OpenAI added 12-hour prompt caching to its API. In OpenAI's words, "12-hour cache writes cost 2.5x the input token price", against 1.25x for a 30-minute write. On GPT-6.1 Terra, that is $2.50 per million tokens against $1.25.
The longer write pays off when requests that share the same instructions or documents arrive hours apart. For traffic that repeats within half an hour, the cheaper 30-minute write is still the better buy. OpenAI's API pricing page lists both write rates for each model.
Gemini 4 Argon: $2/$10 Now, $4/$20 From 1 December
Google made Gemini 4 Argon generally available on 13 Oct 2026. It is Google's flagship reasoning model, with a 1M-token context window.
Until 30 Nov it costs an introductory $2/$10, with cached input at $0.10. From 1 Dec 2026 its list price of $4/$20 applies, with cached input at $0.20.
Argon supports implicit caching only, and it is not on Google's free tier. At its introductory price, it matches Claude Sonnet 5.5 and GPT-6.1 Sol. From December it costs the same as Claude Opus 5.5, and our standard workload rises from $1,100 to $2,200 a month.
If you are testing Argon now, budget at the December rate. Our Gemini 4 Argon pricing guide covers what it does and who it suits.
Gemini 3.5 Flash Now Routes to 3.6 Flash
On 8 Oct, Google deprecated Gemini 3.5 Flash and began routing its API requests to Gemini 3.6 Flash, its replacement. Gemini 3.6 Flash costs an introductory $0.75/$3.75 through 31 Dec 2026. Its $1.50/$7.50 list price returns on 1 Jan 2027.
Our Google Gemini API pricing page has every Gemini rate.
Mistral Large 4 Price Cut and Two Retirements
Mistral launched Mistral Large 4 on 6 Oct 2026 in public preview. It is a multimodal model with a 1M-token context window. It opened with a two-week launch discount, and a $1.36/$4.18 list price after it.
By 30 Oct, Mistral's API pricing page listed Large 4 at $0.70/$2.10, with cached input at $0.07. Mistral did not date the change.
The Mistral Large 4 price cut is roughly half on both input and output. It takes our standard workload from $617 to $315 a month.
Large 3 and Medium 3.5 Retire in April
On 31 Oct, Mistral deprecated Mistral Large 3 and Mistral Medium 3.5. Its models page lists 30 Apr 2027 as the retirement date for both, with Mistral Large 4 as the alternative. Large 3 costs $0.50/$1.50 and Medium 3.5 costs $1.50/$7.50, both with a 256K context window.
For Medium 3.5 users, the move saves money. Large 4 is cheaper on both rates and has about four times the context.
For Large 3 users it costs more. Large 4's rates are 40% higher on both input and output, so the standard workload rises from $225 to $315. You have six months to test it.
See our Mistral API pricing page for every rate.
Smaller Changes That Still Move a Bill
xAI Lists a 20% Batch Discount for Grok 4.7
Our October roundup said Grok 4.7 had no batch discount. By 26 Oct, xAI's pricing page listed Grok 4.7, Grok 4.6 and Grok 4.5 among the models that get 20% off standard rates on its Batch API.
That takes Grok 4.7 from $2/$6 to $1.60/$4.80 for work that can wait. Our standard workload drops from $900 to $720. The discount is well short of the 50% that Anthropic's Batch API gives.
See our xAI Grok API pricing page for every Grok rate.
Meta Raises the Muse Spark Contributor Cap
Meta now caps Muse Spark Contributor mode at 150 requests a minute, up from 100. Contributor mode bills $0.10/$0.20, against $1.25/$4.25 in standard mode.
In return, Meta may use your content to train future Meta models. The higher cap makes it easier to prototype on, but the training terms have not changed.
Subscription Credits That Offset API Spend
Two subscription changes cut what some teams pay for the API. On 7 Oct, Anthropic said Claude Max, Ultra and Team plans now include monthly API credits. Max includes $100 a month and Ultra $200, as our Claude plans guide shows.
On 28 Oct, OpenAI said eligible ChatGPT Pro and Business subscribers can claim "up to $500 per month in free API credits". Our ChatGPT plans guide lists each plan's credits.
Two Corrections to Our Own Data
We fixed two of our own listings. GPT-5.5 Pro bills $60/$270 for prompts over 272K input tokens, which our table had wrong. Gemini 3.1 Flash-Lite became generally available on 7 May 2026, not 19 May. Neither was a vendor price change.
AI API Pricing November 2026: The Rate Card
All prices are USD per million tokens at standard rates, for short prompts and before batch discounts.
| Model | Input | Output | Cached Input | Change Since 5 October |
|---|---|---|---|---|
| Claude Fable 5.1 | $10.00 | $50.00 | $0.25 | None |
| GPT-6 Astra | $10.00 | $50.00 | $1.00 | None |
| Claude Opus 5.5 | $4.00 | $20.00 | $0.20 | None |
| GPT-5.6 Sol | $4.00 | $20.00 | $0.40 | None (promotional, list $5/$30) |
| Gemini 4 Argon | $2.00 | $10.00 | $0.10 | New (13 Oct), $4/$20 from 1 Dec |
| Claude Sonnet 5.5 | $2.00 | $10.00 | $0.10 | Cache reads halved (7 Oct) |
| GPT-6.1 Sol | $2.00 | $10.00 | $0.10 | None |
| GPT-5.6 Terra | $2.00 | $12.00 | $0.20 | None |
| Grok 4.7 | $2.00 | $6.00 | $0.50 | 20% batch discount listed |
| GPT-6.1 Terra | $1.00 | $5.00 | $0.05 | New (26 Oct) |
| Claude Haiku 4.5 | $1.00 | $5.00 | $0.10 | Replaced by Haiku 5.5 |
| Gemini 3.8 Flash | $0.75 | $3.75 | $0.075 | None (introductory until 31 Dec) |
| Mistral Large 4 | $0.70 | $2.10 | $0.07 | New (6 Oct), cut by 30 Oct |
| DeepSeek V4.1 Flash (peak) | $0.30 | $1.20 | $0.006 | None |
| Claude Haiku 5.5 | $0.10 | $0.50 | $0.01 | New (7 Oct), up to 100K tokens |
| GPT-6 Luna | $0.10 | $0.50 | $0.01 | None |
Compare every tracked model side by side in our AI API comparison table.
What the Same Monthly Workload Costs Now
We use the same job every month: 100,000 requests averaging 3,000 input and 500 output tokens. That is 300 million input and 50 million output tokens. Every request is far under 100,000 tokens, so Haiku 5.5 bills its lower rate. At November's rates, the job costs:
- GPT-6 Astra or Claude Fable 5.1: $5,500
- Claude Opus 5.5 or GPT-5.6 Sol: $2,200, with Sol at $3,000 if its promotion ends
- GPT-5.6 Terra: $1,200
- Claude Sonnet 5.5, GPT-6.1 Sol or Gemini 4 Argon: $1,100, with Argon rising to $2,200 on 1 Dec
- Grok 4.7: $900, or $720 on the Batch API
- GPT-6.1 Terra or Claude Haiku 4.5: $550
- Gemini 3.8 Flash: $412.50, rising to $825 on 1 Jan 2027
- Mistral Large 4: $315
- DeepSeek V4.1 Flash: $150 at peak, $75 off-peak
- Claude Haiku 5.5 or GPT-6 Luna: $55
The month's story is in the middle of that list. GPT-6.1 Terra runs the job for $550, half GPT-6.1 Sol's $1,100 and less than half GPT-5.6 Terra's $1,200. At the bottom, Claude now matches OpenAI's $55.
Caching changes the API pricing picture again, and it rewards the models with the cheapest cache reads.
With 80% of input served from cache, Claude Sonnet 5.5 and GPT-6.1 Sol both land at $644, GPT-6.1 Terra at $322, and Claude Haiku 5.5 and GPT-6 Luna at about $33 each, before cache writes. Model your own numbers in the AI pricing calculator.
October's Watch List: What Happened
Last month's roundup flagged these dates. Here is where each one stands.
o4-mini on OpenAI, 23 Oct: Done. OpenAI shut down the o4-mini API on schedule, and we removed it from our tables. OpenAI's recommended replacement is GPT-5.6 Terra.
o4-mini on Azure, 19 Nov: Still on track. Microsoft's retirement schedule lists 19 Nov 2026 for o4-mini, with GPT-5.6 Terra as the replacement.
GPT-5.6 Sol's promotional floor, 21 Nov: Unchanged. OpenAI's API pricing page still says the promotional $4/$20 is "available at least through November 21, 2026". Its list price is $5/$30.
Claude Haiku 4.5, "not sooner than" 15 Oct: That date has passed with no deprecation notice. Anthropic's deprecations table still lists Haiku 4.5 as active.
GPT-5 nano, 11 Dec: Still on track. OpenAI's deprecations page lists 11 Dec 2026, with GPT-5.6 Luna as the recommended replacement.
Google's Flash introductory pricing, 31 Dec: Unchanged. Gemini 3.6 Flash and 3.8 Flash stay at $0.75/$3.75 until 31 Dec, then return to $1.50/$7.50.
Would the GPT-6 family keep pulling prices down? Yes. GPT-6.1 Terra arrived at half GPT-5.6 Terra's input price, and Claude Haiku 5.5 matched GPT-6 Luna's rate on short prompts.
Fun Fact: Three Labs, One Price, Until 30 November
For the rest of November, three different labs sell a model at exactly $2/$10 per million tokens: Claude Sonnet 5.5, GPT-6.1 Sol and Gemini 4 Argon. All three charge $0.10 for cached input too. On 1 December, Argon's price doubles and the three-way tie ends.
What to Watch in November and December
19 Nov: Azure retires o4-mini.
21 Nov: GPT-5.6 Sol's promotional rate reaches its floor date. If OpenAI ends it, Sol returns to $5/$30, and GPT-6.1 Sol at $2/$10 is the obvious model to test.
30 Nov: Claude Sonnet 4.5 retires from the Claude API, with Sonnet 5.5 as the recommended replacement.
Anthropic also removes fast mode for Claude Opus 4.8 that day and points users to Opus 5.5, whose fast mode costs $8/$40. Gemini 4 Argon's introductory rate ends too.
1 Dec: Gemini 4 Argon's $4/$20 list price begins.
11 Dec: OpenAI shuts down GPT-5 nano and other older GPT-5 and o3 snapshots.
31 Dec: Google's introductory Flash pricing ends.
14 Jan 2027: OpenAI removes GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini from its API, with GPT-6 Luna as the recommended replacement for all four. It announced this on 14 Oct, and the details are on its deprecations page.
Frequently Asked Questions
Did Any AI API Prices Go Up in October 2026?
Not after our 5 October roundup. Between 6 and 31 October we logged no list-price rise on an existing model.
That roundup recorded one earlier rise, Llama 4 Maverick's move from $0.20/$0.80 to $0.27/$0.85 on 1 Oct, after its reference host retired it. The next scheduled rise is Gemini 4 Argon, from $2/$10 to $4/$20 on 1 Dec.
What Is the Cheapest Claude Model in November 2026?
Claude Haiku 5.5, at $0.10/$0.50 per million tokens for prompts up to 100,000 tokens, or $0.05/$0.25 on the Batch API. A prompt over 100,000 tokens bills the whole request at $0.50/$2.50.
Is GPT-6.1 Terra Cheaper Than GPT-6.1 Sol?
Yes, by half. GPT-6.1 Terra costs $1/$5 with cached input at $0.05, and GPT-6.1 Sol costs $2/$10 with cached input at $0.10. Above 272K input tokens, Terra bills $2/$7.50 and Sol $4/$15.
When Does Gemini 4 Argon's Price Go Up?
On 1 Dec 2026. The introductory $2/$10 runs through 30 Nov, and the $4/$20 list price, with cached input at $0.20, applies from 1 Dec.
When Do Mistral Large 3 and Medium 3.5 Retire?
On 30 Apr 2027, according to Mistral's models page. Mistral deprecated both on 31 Oct 2026 and names Mistral Large 4 as the alternative.
That completes our AI API pricing November 2026 review as the month begins. The list above this review logs each change as it lands, and our pricing history explorer shows how every model got here.
