Release tracker / Anthropic
Anthropic: latest announcements and model releases
As of 20 November 2026, Anthropic's latest model release is Claude Mythos 5.2, released on 9 Nov 2026 (11 days ago). Claude Fable 5.2 came out the same day. We have recorded 27 Anthropic launches since 4 Mar 2024.
Every Anthropic model launch we have recorded, newest first: Fable, Haiku, Sonnet, Opus. Each one links to its source, and to how it ranks on our benchmarks where we have tested it.
- Latest
- Claude Mythos 5.2, 9 Nov 2026
- Launches
- 27
- Since
- 4 Mar 2024
- Since last launch
- 11 days
Anthropic's latest model
-
Claude Mythos 5.2
No results in our own benchmarks yet.
Claude Mythos 5.2 came out 11 days ago. Anthropic posts its own announcements at www.anthropic.com/news. Spotted a launch we're missing? Tell us.
What Anthropic may ship next
Each line's last launch and when its next model is likely, sorted by the likeliest date. Hover or tap a row for the reasons.
No launches look likely this week. The next is Claude Opus 5.6+, around 15 Dec.
How our recent estimates did
- Claude Fable 5.2 launched 9 Nov. On 8 Nov we had Claude Fable 5.2+ at 10–21 Nov, around 12 Nov: 1 day before our window.
Our estimate, from prediction markets including Polymarket and Manifold, each line's usual gap between launches and pre-release sightings. These are betting markets: a signal, not a forecast. Updated 19 Nov 2026.
How often Anthropic ships
Lines with at least three launches in the last two years. Amber: longer since the last launch than the line's average gap. Grey: no launch for 90 days or more, marked quiet.
| Line | Size | Launches | Average gap | Latest | Last launch |
|---|---|---|---|---|---|
| Opus | Large | 8 | 70 days | Claude Opus 5.522 Sep 2026 | 59 days ago |
| Sonnet | Medium | 6 | 116 days | Claude Sonnet 5.528 Sep 2026 | 53 days ago |
| Fable | Large | 4 | 72 days | Claude Fable 5.29 Nov 2026 | 11 days ago |
Every Anthropic release, with its release date
-
Claude Mythos 5.2
No results in our own benchmarks yet.
-
Claude Fable 5.2 came first on our survey analysis and ad budget benchmarks and third on product listings and slide decks, a large step up from Fable 5.1. It also cost about ten times as much per task as GPT-6.1 Sol.
Our test results: CatalogBench · DeckBench · ROASBench · SurveyBench. Read our review →
#1 of 39 overall$8.00 in · $40.00 out per MProvider source
-
We ran Claude Haiku 5.5 through our business benchmarks and 18 playground tests on launch day. It is a big step up from Haiku 4.5 and very cheap, but GPT-6 Luna, at the same price, beat it on listings, decks and ad planning.
Our test results: CatalogBench · DeckBench · ROASBench · SurveyBench. Read our review →
#27 of 39 overall$0.10 in · $0.50 out per MProvider source
-
Our test results: BulletBench · CatalogBench · DeckBench · ROASBench.
#14 of 39 overall$2.00 in · $10.00 out per MProvider source
-
Our test results: CatalogBench · DeckBench · ROASBench · SurveyBench.
#4 of 39 overall$4.00 in · $20.00 out per MProvider source
-
Our test results: CatalogBench · DeckBench · ROASBench.
#12 of 39 overall$10.00 in · $50.00 out per MProvider source
-
No results in our own benchmarks yet.
$5.00 in · $25.00 out per MProvider source
-
No results in our own benchmarks yet.
$2.00 in · $10.00 out per MEpoch AI source
-
No results in our own benchmarks yet.
$10.00 in · $50.00 out per MProvider source
-
No results in our own benchmarks yet.
$5.00 in · $25.00 out per MProvider source
-
No results in our own benchmarks yet.
$5.00 in · $25.00 out per MProvider source
-
Claude Mythos Preview
No results in our own benchmarks yet.
-
No results in our own benchmarks yet.
$3.00 in · $15.00 out per MProvider source
-
No results in our own benchmarks yet.
$5.00 in · $25.00 out per MProvider source
-
No results in our own benchmarks yet.
$5.00 in · $25.00 out per MProvider source
-
Our test results: BulletBench · CatalogBench · DeckBench · ROASBench · SurveyBench.
#39 of 39 overall$1.00 in · $5.00 out per MProvider source
-
No results in our own benchmarks yet.
$3.00 in · $15.00 out per MProvider source
-
No results in our own benchmarks yet.
$15.00 in · $75.00 out per MProvider source
-
No results in our own benchmarks yet.
-
No results in our own benchmarks yet.
-
No results in our own benchmarks yet.
-
Claude 3.5 Sonnet (2024-10-22)
No results in our own benchmarks yet.
-
No results in our own benchmarks yet.
-
No results in our own benchmarks yet.
-
No results in our own benchmarks yet.
-
No results in our own benchmarks yet.
-
No results in our own benchmarks yet.
Our reviews of Anthropic models
The launch reviews we have published for Anthropic models, the newest release first.
- Claude Fable 5.2 review: first on our survey and ad budget tests, and one of the dearest models to run Claude Fable 5.2 · 10 Nov 2026
- Claude Haiku 5.5 Is Far Better Than Haiku 4.5 at a Tenth of the Price. GPT-6 Luna Still Beats It at Business Work. Claude Haiku 5.5 · 8 Oct 2026
Questions
- What is Anthropic's latest model?
- Claude Mythos 5.2, released 9 Nov 2026 (11 days ago). It is the newest of the 27 Anthropic launches we have recorded since 4 Mar 2024.
- When will Anthropic release its next model?
- Anthropic has not announced its next model. It usually releases a new Opus model about every 70 days; the last was Claude Opus 5.5, 59 days ago. It usually releases a new Sonnet model about every 116 days; the last was Claude Sonnet 5.5, 53 days ago. It usually releases a new Fable model about every 72 days; the last was Claude Fable 5.2, 11 days ago. Our release signal, which draws on prediction markets, puts the next Claude Opus 5.6+ likely 3–29 Dec, around 15 Dec (medium confidence); it is a signal, not a forecast we stand behind.
- How often does Anthropic release new models?
- Over the last two years: a new Opus model about every 70 days (8 launches); a new Sonnet model about every 116 days (6 launches); a new Fable model about every 72 days (4 launches).
Other providers: OpenAI, Mistral AI, Alibaba, Google, DeepSeek, Meta, xAI, Z.ai, Moonshot AI, Microsoft, MiniMax, NVIDIA, Xiaomi, Amazon, StepFun, Thinking Machines. Or see every launch on one timeline.
A launch is the first day a model was publicly available, previews included; a family's sizes released together count once, and later dated updates of a model are left out. Main lines are checked by hand against the provider's own changelog or announcement where we could read it ("Provider"), otherwise a reference page ("Reference"); the rest come from Epoch AI's model data (CC BY 4.0).