Claude Haiku 5.5 status tracker: released on 2026-10-07
As of 2026-10-26, Claude Haiku 5.5 has been out since 2026-10-07: model ID claude-haiku-5-5, 1M context, 128K max output, and two price rows by prompt length. Prompts up to 100,000 tokens pay $0.10 input / $0.50 output per million tokens; longer prompts pay $0.50 / $2.50. As of 2026-10-26, claude-haiku-5-5 is on sale and callable on QCode's /models; billing is per token and per-model rates are listed on /models.
Updated 2026-10-30
One key across Claude, GPT and Chinese models, billed per token.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Four key points
Official release date
The official Haiku 5.5 model page and the matching Claude Platform release-notes entry both carry October 7, 2026, and the notes log the launch the same day. Before that, Anthropic had named it as part of the 5.5 family twice (2026-09-22 and 2026-09-28), without giving a date.
Input / output (prompts up to 100K)
For prompts up to 100,000 tokens: $0.10 input, $0.50 output and $0.01 cache reads per million tokens. Prompts over 100,000 tokens use the second row: $0.50 input, $2.50 output, $0.05 cache reads. Official wording: a request whose prompt is over 100,000 tokens pays higher prices.
Context window / max output
Official model page: 1M token context window, 128K max output tokens, adaptive thinking, default effort medium. Haiku 4.5 is 200K / 64K.
Haiku 4.5 lifecycle status
As of 2026-10-26, the deprecations table still lists claude-haiku-4-5-20251001 as Active with Not sooner than October 15, 2026; that date has passed, but it is a floor, not a shutdown, and the deprecation history has no retirement notice for it. The Haiku column of the models overview lists Claude Haiku 5.5 with the Claude API ID claude-haiku-5-5.
What has now been published
As of 2026-10-26 Haiku 5.5 is released, and the four items that were missing are now official: release date 2026-10-07; model ID claude-haiku-5-5 (anthropic.claude-haiku-5-5 on Amazon Bedrock); 1M context and 128K max output; and prices in two rows by prompt length. Anthropic positions it for high-volume, latency-sensitive tasks such as classification, extraction, and routing. Against Haiku 4.5 ($1 / $5, 200K / 64K) the short-prompt row is cheaper and the window is larger, but the docs also say the same text counts as approximately 30% more tokens on Haiku 5.5, so re-estimate costs with the new tokenizer.
2026-10-07: from "coming weeks" to launch
The Opus 5.5 launch page of 2026-09-22 first listed Sonnet 5.5 and Haiku 5.5 as part of the 5.5 family, and the Sonnet 5.5 launch page of 2026-09-28 supplied Haiku 5.5's positioning (built for high-volume and cost-sensitive applications); neither gave a release date. On 2026-10-07 it shipped. The Claude Platform release notes read: We've launched Claude Haiku 5.5 (claude-haiku-5-5), our most capable model tuned for high-volume and latency-sensitive work. Anthropic says it costs around 75% less to run than Haiku 4.5 on average; that is a vendor claim. The same day Anthropic halved the price of Sonnet 5.5 cache reads.
Public timeline
On 2026-09-22 the Opus 5.5 launch page first listed Sonnet 5.5 and Haiku 5.5 as part of the 5.5 family. It is the first time Haiku 5.5 appears in Anthropic's 5.5-family wording: no date and no price attached. Haiku 5.5 released on 2026-10-07.
On 2026-09-28 Sonnet 5.5 shipped, and its launch page supplied Haiku 5.5's role line (high-volume and cost-sensitive applications) but still gave no release date; Haiku 5.5 released on 2026-10-07.
2026-10-07: Haiku 5.5 launches with model ID claude-haiku-5-5, 1M context and 128K max output; the Haiku column of the models overview switches from Haiku 4.5 to Haiku 5.5. Rechecked on 2026-10-26: claude-haiku-4-5-20251001 is still Active.
Confirmed vs unverified
Confirmed (official wording)
As of 2026-10-26 all of this can be matched word for word on official pages: Haiku 5.5 was released on 2026-10-07, the day of the release-notes entry that opens with We've launched Claude Haiku 5.5 (claude-haiku-5-5), with Claude API model ID claude-haiku-5-5; 1M token context, 128K max output, adaptive thinking, default effort medium. Prices come in two rows: for prompts up to 100,000 tokens, $0.10 input, $0.50 output, $0.125 for 5-minute cache writes, $0.20 for 1-hour cache writes and $0.01 cache reads; for prompts over 100,000 tokens, $0.50, $2.50, $0.625, $1 and $0.05 (per million tokens). Status Active, retirement not sooner than 2027-10-07. Migration: manual extended thinking (budget_tokens) returns an error, non-default temperature, top_p or top_k also return an error, and the same text counts as about 30% more tokens. The Claude Code docs require v2.1.293 or later with Haiku 5.5. Haiku 4.5 remains Active with Not sooner than October 15, 2026.
Unverified · nothing carried
As of 2026-10-26, two claims have no official basis or are vendor statements only. 1) "Haiku 4.5 shuts down on 2026-10-26": the retirement column reads Not sooner than October 15, 2026, a floor, and the deprecation history has no notice for it as of the check date. 2) Benchmarks and cost: the launch page carries Anthropic's own results and an "around 75% less to run" claim, both vendor statements, not third-party measurements.
Two Haiku generations
Haiku 4.5 vs Haiku 5.5 (both published)
Haiku 4.5: $1 input, $5 output, $0.10 cache reads, 200K context, 64K max output, claude-haiku-4-5-20251001 is marked "Retired" on QCode's /models. Haiku 5.5: for prompts up to 100,000 tokens $0.10 input, $0.50 output, $0.01 cache reads; above that $0.50 / $2.50 / $0.05; 1M context, 128K max output. Converted from the official price table, the short-prompt input price is one tenth of Haiku 4.5 ($0.10 vs $1) and the long-prompt row is half ($0.50 vs $1); since the same text yields about 30% more tokens, recompute the gap on your own traffic.
Official deprecation wording vs a shutdown date
The deprecations table gives Haiku 4.5 Not sooner than October 15, 2026: a floor, not a shutdown on October 15. The same page says publicly released models get at least 60 days' notice before retirement. As of 2026-10-26 the deprecation history has no notice for Haiku 4.5, so by that notice period it cannot retire earlier than 60 days after a notice is sent. Since Haiku 5.5 shipped, the Haiku column of the models overview lists Claude Haiku 5.5. For comparison, Sonnet 4.5 got its notice on 2026-09-30, with retirement on the Claude API set for 2026-11-30.
How to choose today
As of 2026-10-26 there are three routes. 1) Moving to Haiku 5.5: on the official Claude API, switch the model ID to claude-haiku-5-5; QCode sells the model on /models; set the model to claude-haiku-5-5 with the same key and endpoint (https://api.qcode.cc/api); billing is per token and per-model rates are listed on /models. Either way, follow the official migration guide: change budget_tokens thinking to adaptive, remove temperature, top_p and top_k, end messages with a user turn, and recount tokens and max_tokens (the same text is about 30% more tokens). 2) Not migrating yet: Haiku 4.5 (claude-haiku-4-5-20251001) has an official price of $1 input / $5 output and 200K context; as of 2026-10-26 it is marked "Retired" on QCode's /models. 3) Stronger reasoning than Haiku: on QCode, use Sonnet 5.5 (claude-sonnet-5-5), official price $2 / $10, 1M context and 128K max output. When estimating Haiku 5.5 costs, remember that any prompt over 100,000 tokens is priced on the $0.50 / $2.50 row, so long-document jobs should not be budgeted at the entry price. Full specs and the change-by-change migration walkthrough are on the Claude Haiku 5.5 complete guide at /claude-haiku-5-5-guide.
QCode: Haiku 5.5 now supported
As of 2026-10-26, claude-haiku-5-5 is on sale and callable on QCode's /models, in three steps: 1) sign up, top up with Alipay, WeChat Pay, a Visa / Mastercard card, Apple Pay, Google Pay or cryptocurrency, and create a key starting with cr_ in the dashboard; 2) set ANTHROPIC_BASE_URL=https://api.qcode.cc/api (Anthropic protocol, no trailing slash) and ANTHROPIC_AUTH_TOKEN to your cr_ key; 3) set the model to claude-haiku-5-5 in your API request. Claude Code requires v2.1.293 or later with Haiku 5.5. claude-haiku-4-5-20251001 (Haiku 4.5) is marked "Retired" on QCode's /models; when you need stronger reasoning, claude-sonnet-5-5 is callable on QCode. Before switching from Haiku 4.5, adjust parameters per the official migration guide (the same text is about 30% more tokens). The full walkthrough is on /claude-haiku-5-5-guide. Want to try Haiku 5.5 for a few days without a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days; if you already have a Personal Plan, add a booster pack when you need more quota. Billing is per token; per-model rates are listed on /models.
Questions
When was Haiku 5.5 released?
It is out: the official Haiku 5.5 model page and the Claude Platform release-notes entry both carry October 7, 2026, and the notes open with We've launched Claude Haiku 5.5 (claude-haiku-5-5). Before that, Anthropic had named it as part of the 5.5 family twice (2026-09-22 and 2026-09-28), both times without a date.
How much does Haiku 5.5 cost?
Two rows by prompt length, per million tokens: for prompts up to 100,000 tokens, $0.10 input, $0.50 output, $0.01 cache reads, $0.125 for 5-minute and $0.20 for 1-hour cache writes; for prompts over 100,000 tokens, $0.50 input, $2.50 output, $0.05 cache reads, and $0.625 / $1 cache writes. The pricing page states that a request whose prompt is over 100,000 tokens pays higher prices. The threshold counts all of a request's input tokens, cache reads and cache writes included, and each request is priced on its own: a request over the threshold pays the higher prices for the whole request even when part of its prompt is a cache hit. For comparison, Haiku 4.5 is $1 input and $5 output.
What are the context window and max output?
A 1M token context window and 128K max output tokens. Haiku 4.5 is 200K / 64K; Sonnet 5.5 and Opus 5.5 are both 1M / 128K. Note that Haiku 5.5 prices depend on prompt length: prompts over 100,000 tokens use the higher row, so long contexts cost more than short prompts.
Can I call Haiku 5.5 through QCode?
Yes. As of 2026-10-26, claude-haiku-5-5 is on sale on QCode's /models: use the same key on the same Anthropic-compatible endpoint (https://api.qcode.cc/api) and set the model to claude-haiku-5-5; the Claude Code docs require v2.1.293 or later with Haiku 5.5. Billing is per token and per-model rates are listed on /models. claude-haiku-4-5-20251001 is marked "Retired" on QCode's /models; for stronger reasoning, claude-sonnet-5-5 is also callable, at a higher official price and with parameters changed per the official migration guide. On the official Claude API the Haiku 5.5 model ID is also claude-haiku-5-5.
Will Haiku 4.5 be retired?
As of 2026-10-26, Haiku 4.5 is still Active with no retirement scheduled. The retirement column reads Not sooner than October 15, 2026; that date has passed, but it is a floor, not a shutdown, and the deprecation history has no notice for it. Anthropic commits to at least 60 days' notice before retiring a publicly released model, so the earliest retirement is 60 days after a notice. The official launch page calls Haiku 5.5 the cheapest, fastest, and most capable small model Anthropic has released; whether to migrate is yours to test. These dates cover Anthropic-operated platforms such as the Claude API; Amazon Bedrock and Google Cloud set their own schedules.
How do I pay for QCode and start calling Haiku 5.5?
Top up with a credit card (Visa, Mastercard), Apple Pay, Google Pay, Alipay, WeChat Pay or cryptocurrency (several coins and networks), choose a plan and create a key starting with cr_ in the dashboard, then set the model to claude-haiku-5-5. It is fully self-serve and takes a few minutes. Billing is per token; per-model rates are listed on /models.
Sources
Fetched and checked on 2026-10-26: the Claude Platform release notes (2026-10-07 Haiku 5.5 launch entry), the Claude Haiku 5.5 model page (release date, model IDs, both price rows, context and max output), the official pricing page (Haiku 5.5 rows and the prompt-length pricing note), the Claude Platform models overview, the model deprecations page (Haiku 4.5 Active, Not sooner than October 15, 2026, at least 60 days' notice), the Haiku 5.5 migration guide, the What's new and prompting pages, and the Anthropic Haiku page (which compares Haiku 4.5 with Haiku 5.5), the Opus 5.5 (2026-09-22) and Sonnet 5.5 (2026-09-28) launch pages, and the QCode /models and /claude-haiku-5-5-guide pages, plus the QCode endpoint documentation. Rechecked on 2026-10-26: the model deprecations page (Haiku 4.5 still Active, no retirement notice), the Claude Code model configuration docs (Haiku 5.5 effort levels and default), and QCode's environment-variable and payment-method docs.
Haiku 5.5 is callable on QCode: set model to claude-haiku-5-5
As of 2026-10-26, Haiku 5.5 (claude-haiku-5-5) and claude-sonnet-5-5 are on sale and callable on QCode's /models; use the QCode endpoint and your key, and set model to the ID you want. Fully self-serve and set up in minutes; billing is per token and per-model rates are listed on /models.
Related
Claude Haiku 5.5 complete guide
Two price tiers ($0.10 / $0.50 up to 100K, $0.50 / $2.50 above), 1M / 128K, platform IDs and how it differs from Haiku 4.5.
Claude Haiku 5.5 breaking changes and errors
The five breaking changes from Haiku 4.5 to Haiku 5.5, each with the official error behaviour and fix.
Claude Opus 5.5 guide
The first model of the 5.5 family to ship: Anthropic wrote «in the coming weeks» for the first time on this release page.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
QCode is not affiliated with Anthropic. This page separates published information (Haiku 5.5 release date, model ID, prices and context; Haiku 4.5 status and retirement floor) from vendor claims (cost statements) and carries no rumors. Official pages were checked on 2026-10-26, with the deprecations table and Claude Code docs rechecked on 2026-10-26; the official pages prevail, and model availability is shown on /models.