Judging by how often Anthropic starts a new Claude generation and by prediction market odds, Claude 6 is most likely to come out in the first half of 2027.
We've gathered everything we know about Anthropic's next generation. In this article, you'll find the most likely Claude 6 release date and what we know about the models the new generation will include.
The most likely Claude 6 release date is the first half of 2027.
As of this writing, Anthropic hasn't announced Claude 6 or given any indication of when it might launch. Still, we can piece together what we know to estimate a possible release date. Take this with a pinch (or rather, a pincher) of salt, though. These dates are far from set in stone.
In the S-1 filing for its IPO, published on October 26, 2026, Anthropic states:
We make decisions about model development and deployment, including whether to begin training a new model, on a case-by-case basis. … Depending on these assessments, we may delay or suspend training or related research, delay releases, or limit access to models until we have confidence they can be developed and deployed safely.
In other words, Anthropic decides whether to begin training each new model on a case-by-case basis. It may also delay a release or restrict access until it's confident the model can be deployed safely. So even if Anthropic has an internal launch date in mind, there's no guarantee it'll stick to it.
That said, AI companies generally try to maintain a somewhat predictable release schedule to keep users happy and, perhaps more importantly, give investors confidence in their progress. With that in mind, let's look at the release history of previous Claude models and see what it tells us about when Claude 6 might arrive.
Anthropic launches a new Claude generation about once a year:
| Generation | First models | Release date | Gap |
|---|---|---|---|
| Claude 3 | Claude 3 Opus, Claude 3 Sonnet | March 4, 2024 | — |
| Claude 4 | Claude Opus 4, Claude Sonnet 4 | May 22, 2025 | 14.5 months |
| Claude 5 | Claude Fable 5, Claude Mythos 5 | June 9, 2026 | 12.5 months |
| Claude 6 | — | Expected in the first half of 2027 | — |
The gap between generations keeps getting shorter, so if the next gap is the same as or shorter than the last, we can place the Claude 6 release date somewhere in June 2027.
Between generations, Anthropic releases point updates every two to three months. The Claude 4 generation had five point updates before Claude 5 arrived: Claude 4.1, 4.5, 4.6, 4.7 and 4.8. Claude 5 has had two so far — Claude 5.1 in September and Claude 5.5 in September and October — so there's room for more 5.x updates before Claude 6.
Frequent updates also make business sense for Anthropic. According to the aforementioned S-1, models released in the six months before September 30, 2026 brought in over 80% of revenue in the third quarter. New versions quickly take over most of the usage, which gives Anthropic a reason to release often.
Prediction markets also expect Claude 6 in the first half of 2027. At the time of writing, traders on Polymarket and Manifold give these odds of a Claude 6 release:
| Market | Released by | Odds |
|---|---|---|
| Manifold | December 1, 2026 | 17% |
| Manifold | January 1, 2027 | 24% |
| Manifold | March 1, 2027 | 56% |
| Polymarket | March 31, 2027 | 39% |
| Polymarket | June 30, 2027 | 73% |
| Polymarket | December 31, 2027 | 89% |
Manifold points to an earlier date because while Polymarket only counts a public release of a Claude 6 model, like Claude Opus 6 or Fable 6, and ignores a Mythos 6 that's available only to vetted organizations, Manifold counts any Claude 6 model, including a limited preview.
The public launch of Claude 5 came after exactly such a preview. Anthropic gave Claude Mythos Preview to a small group of companies in April 2026, two months before Claude Fable 5 released to the general public. A similar Mythos 6 preview could appear well before the public Claude 6 models.
The next Claude model will most likely be another Claude 5 update. In September 2026, users on X shared demos from a model they called Claude Fable 5.2, which Anthropic appeared to be testing on some Claude Fable 5.1 requests. Claude Fable 5.2 hasn't been released yet, and it's the most likely candidate for Anthropic's next release.
Anthropic updates Fable and Mythos separately from the other Claude models. For example, Claude Fable 5.1 came out on September 1, 2026, three weeks before Claude Opus 5.5. So Fable 5.2 can arrive without new versions of Opus or Sonnet.
Since Claude 3, every Claude generation has had Opus, Sonnet and Haiku models, from the most capable to the fastest and cheapest. Claude 5 added a fourth class above Opus, called Mythos. If Anthropic keeps this structure, Claude 6 will include Claude Fable 6 and Claude Mythos 6 at the top, followed by Claude Opus 6, Sonnet 6 and Haiku 6.
The models of one generation don't come out on the same day. Claude 5 started with Claude Fable 5 and Mythos 5 on June 9, 2026, and the other models followed over the next weeks and months. Claude 6 will most likely roll out the same way, so the first model with a "6" in its name may well be Fable 6 or Mythos 6.
Claude Opus is Anthropic's main model for hard tasks like agentic coding and long documents. The current version, Claude Opus 5.5, came out on September 22, 2026. Anthropic built it as an everyday model for companies that powers agents running whole projects on their own.
The Claude Opus 6 release date will most likely fall a few weeks after the first Claude 6 models. In the Claude 5 generation, Opus 5 came out on July 24, 2026, about six weeks after Fable 5.
Claude Sonnet is the default model on the free Claude plan, so it's the version of Claude that most free users chat with. Sonnet 5 came out three weeks after the first Claude 5 models, so Claude Sonnet 6 is likely to follow the Claude 6 launch closely too.
Claude Haiku is the smallest and cheapest Claude model, and it gets updates less often. There was no Haiku 5 at all: the next Haiku model, Haiku 5.5, arrived only in October 2026. Claude Haiku 6 may likewise come only with a later Claude 6.x update.
Fable and Mythos are two versions of the same model, which Anthropic calls its most capable model class. Claude Fable is available to everyone, but with safety restrictions: when a request touches a high-risk area like cybersecurity or biology, Fable hands it to a less capable model. For example, Claude Fable 5 passed such requests to Claude Opus 4.8. Claude Mythos has looser restrictions and is available only to vetted organizations working in cybersecurity and life sciences, through programs like Project Glasswing.
The top Claude 6 model will most likely come in the same two versions, Claude Fable 6 and Claude Mythos 6. Its public release can also depend on regulators. In June 2026, the US Department of Commerce restricted Fable 5 and Mythos 5 three days after launch, and Anthropic switched both models off for two to three weeks, until it updated its safeguards and brought them back.
Anthropic doesn't say where it trains each model, but we can once again turn to the S-1 form to fish out some interesting facts. The company has secured compute for research and model development. It trains and runs Claude on four families of chips, including Amazon Trainium and Google TPUs, and moves workloads between them to get the best price and performance.
To understand the scale of Claude 6 training, you need to know about Anthropic's deals with the big cloud providers.
Amazon, Anthropic's main training partner, runs the Project Rainier campus in New Carlisle, Indiana, which came online in October 2025 with more than 500,000 Trainium2 chips working on Claude. In April 2026, Anthropic signed a new agreement with Amazon and committed to spend at least $110 billion on AWS between May 2026 and April 2036. The deal gives Anthropic up to 5 GW of Trainium capacity, and about 1 GW of Trainium2 and Trainium3 chips should be running by the end of 2026.
Google is Anthropic's second big source of compute. In October 2025, Anthropic secured access to up to one million Google TPUs, with well over 1 GW of capacity in 2026. A new agreement signed in March 2026 commits Anthropic to spend about $114.5 billion on Google Cloud through July 2033.
Google and Broadcom are also building several gigawatts of next-generation TPUs for Anthropic, which start coming online in 2027. Anthropic has also agreed to buy more TPU capacity for 2028–2031, which it estimates at $125–150 billion.
Anthropic also uses Microsoft's cloud under agreements signed in February and October 2026: it committed to spend at least $44.9 billion on Azure between November 2026 and May 2033.
Compute limits how fast Anthropic can train and ship new models. Anthropic has "consistently operated in a compute-constrained environment where customer demand exceeds available compute". A large part of the new capacity starts arriving at the end of 2026 and in 2027, which fits a Claude 6 trained on it and released in 2027.
Claude also helps build Claude. Anthropic reports that Claude has helped write the code that makes running its models more efficient, and that this has sped up model development and release timelines.
Anthropic hasn't revealed any Claude 6 features yet, but the evolution of the Claude 5 models, along with new information the company disclosed ahead of its IPO, gives us a good idea of where Claude is heading. Claude 6 will likely build on those developments.
So, what can we expect from Claude 6?
Anthropic trains Claude more and more for agentic work, where the model takes a task and works on it by itself, writing code and running tools. In the company itself, Claude basically went from suggesting single lines of code to running fleets of agents that work across production codebases on their own. Post-training, the stage where a model learns to write production code and follow multi-step instructions, is now one of Anthropic's largest research investments.
Each update brings a visible jump here. On Terminal-Bench 4.0, a benchmark of real tasks in a computer terminal, Claude Opus 5.5 scores 11 points higher than Opus 5 in Artificial Analysis tests. For Claude 6, this means agents that can work on a codebase for longer and need fewer checks from a developer.
Claude's ability to find and exploit software vulnerabilities grew fast during the Claude 5 generation. On ExploitBench, a benchmark from Carnegie Mellon researchers, Claude's score rose from 40% with Opus 4.8 to 78% with Mythos 5 and 91% with Opus 5.5, all within about four months.
These skills help security teams find and patch vulnerabilities, but attackers can use them too. Before each release, Anthropic's Frontier Red Team checks whether the new model gives attackers meaningful new capabilities, and separate classifiers watch for requests about malware and phishing. If Claude 6 raises these scores again, the results of this review can delay its launch or limit who gets the top model, as happened with Mythos.
Claude Fable 5.1 reduced the cost of typical workloads by about 25% compared to Fable 5, with savings of up to 45% on complex coding and agentic tasks. Claude Opus 5.5 followed a similar trend, cutting the cost per task by roughly 40% compared to Opus 5. Part of that reduction came from lower API prices, which dropped from $5/$25 to $4/$20 per million input/output tokens.
There's also growing pressure from Chinese open-weight models such as Qwen3.8-Max, Kimi K3, and GLM-5.2. These models are getting closer to the capabilities of leading proprietary systems while costing considerably less. And businesses are starting to take notice. As Fortune reported in August 2026, Thomson Reuters adopted a Qwen-based model for document review tasks previously handled by Claude, while legal AI company Harvey built a specialized model on top of Kimi K3.
Given these developments, we expect Anthropic to continue making Claude cheaper to run with the Claude 6 generation, whether through lower API prices, more efficient token usage, or further reductions in prompt caching costs.
There's no reliable Claude 6 leak yet: no Claude 6 model ID has shown up anywhere, and Anthropic hasn't said a word about it. Still, a few rumors have spread in the past months:
Leaked names can also change before release. In February 2026, a Google Vertex AI log showed the model ID claude-sonnet-5@20260203, and many expected Claude Sonnet 5. On February 17, Anthropic released that model as Claude Sonnet 4.6. So even a real "6" in a leaked model ID won't guarantee the name of the final model.
Claude 6 will most likely be available in the same places as the current Claude models.
Claude 6 will be added to Overchat AI as soon as Anthropic releases it. Until then, you can chat with Claude Opus 5.5 and Fable 5.1, Anthropic's best current models, in Overchat AI next to GPT-6 Astra and other models, without a separate Claude subscription.
In the Claude apps, which models you get depends on your plan. On the current Claude plans, the free plan uses Sonnet 5.5 and Pro adds Opus 5.5. Fable 5.1 is included in the Max plans, while Pro users pay for it separately. If Claude 6 follows the same split, free users will get Sonnet 6, while Opus 6 and Fable 6 will need a paid plan.
Developers get new Claude models in the Claude API on release day. Claude models are also available on the big cloud platforms like Amazon Bedrock and Google Cloud Vertex AI, and Claude 6 will most likely be added there too.
Below are short answers to the most common questions about Claude 6.
No, Claude 6 isn't out yet. The newest Claude model is Claude Haiku 5.5, released on October 7, 2026, and the most capable public Claude models are Claude Opus 5.5 and Claude Fable 5.1.
No. Claude Mythos is the class of Anthropic's most capable models, and its versions so far, Mythos 5 and Mythos 5.1, belong to the Claude 5 generation. A Claude Mythos 6 would be part of Claude 6.
Claude Sonnet 6 will most likely be free to use with limits, because Anthropic puts its Sonnet models on the free Claude plan. Opus 6 and Fable 6 will likely require a paid plan.
Probably not. Claude models read text and images and answer with text, and Anthropic doesn't have its own image generation model. Today, you can create images with Claude only through third-party tools.
Claude 6 hasn't been announced, and Anthropic doesn't give release dates in advance. Judging by its yearly cadence and prediction market odds, Claude 6 will most likely come out in the first half of 2027, about a year after Claude 5. Before that, expect more Claude 5 updates, with Claude Fable 5.2 as the most likely next release.
When Claude 6 arrives, it will most likely be a family of models led by Fable 6 and Mythos 6, trained on the Amazon and Google compute that Anthropic is bringing online in 2026 and 2027. Until then, you can use Claude Opus 5.5 and Fable 5.1 in Overchat AI and switch to Claude 6 as soon as it's out.
